AI-Summaries

About:
This study develops a holistic Generative AI (GenAI) Maturity Model to help organizations assess
and improve their GenAI capabilities. Using a design science research approach, the model
combines insights from academic literature and expert interviews to provide a practical,
theoretically-grounded framework for guiding GenAI adoption.

Problem:
While Generative AI is transforming business operations, many organizations struggle to adopt it
effectively. Existing AI maturity models are too general and fail to address the unique challenges
and capabilities required for successful GenAI implementation, leaving a practical gap for managers
seeking guidance.

Outcomes:
- The study presents a preliminary GenAI Maturity Model (GAI-MM) to guide organizations on their
adoption journey.
- The model is structured across five key dimensions: People, Process, Technology, Data, and
Organization.
- It defines five evolutionary maturity levels, from Level 1 (GenAI as a simple Tool) to Level 5
(GenAI as an autonomous part of the Organization).
- Each level provides specific characteristics across the dimensions, offering a clear roadmap for
organizations to benchmark their current state and plan for future development.

Podcast:

About:
This study investigates the key factors that build user trust in generative AI assistants, specifically
within the complex domain of tax law. Using a mixed-methods approach, the research compared
two AI prototypes—one designed with trust-enhancing features and one without—through
quantitative surveys and qualitative interviews with legal experts.

Problem:
While generative AI offers significant potential to streamline work in fields like tax law, its adoption is
hindered by user distrust due to concerns about accuracy, transparency, and reliability. In
high-stakes environments where errors can lead to severe penalties, understanding how to design
trustworthy AI is crucial for realizing its benefits and ensuring safe, effective human-AI
collaboration.

Outcomes:
- Designing AI assistants to be more transparent by citing sources and showing conversation
history significantly increases user trust.
- Features that make the AI more human-like (anthropomorphism), such as using a conversational
tone and providing a personalized greeting, also build trust.
- Users trust an AI more when it is upfront about its capabilities and limitations, aligning with social
and ethical norms.
- A higher level of trust in the AI directly leads to a greater intention to use the tool for professional
tasks.
- Legal experts unanimously preferred the AI prototype that incorporated these trust-building design
features, deeming it more reliable for tax law research.

Podcast:

About:
This conceptual paper analyzes the dual impact of platform-based work on women's careers and
lives. Using case studies of 'mum bloggers', OnlyFans creators, and crowd workers, the study
explores how digital platforms can simultaneously offer empowerment while introducing significant
new risks. The research highlights the 'double-edged sword' nature of this work, demonstrating its
complex relationship with existing gendered power structures.

Problem:
Traditional employment often fails to accommodate the caregiving responsibilities that
disproportionately fall on women, creating significant career barriers. Platform-based work, with its
promise of flexibility, has emerged as an attractive alternative. However, it is unclear whether this
new form of work genuinely empowers women or simply perpetuates old inequalities in a new,
digital guise.

Outcomes:
- Platform work can empower women by providing flexible schedules to balance caregiving with
paid work, fostering financial independence, skill development, and supportive online communities.
- This work carries significant risks, including a lack of social protections (e.g., health insurance,
pensions), unpredictable income streams dependent on opaque algorithms, and financial instability.
- Women on these platforms face heightened mental health challenges from online harassment and
abuse, as well as personal rights risks related to data privacy, security, and physical safety.
- Ultimately, the risks associated with platform work can reinforce existing gender inequalities,
confining women to domestic roles and maintaining their financial dependency, rather than
dismantling patriarchal structures.

Podcast:

About:
This study analyzes the LinkedIn profiles of alumni from 43 major German universities to map the
supply and migration of graduates with entrepreneurial and technical skills. The research identifies
which universities produce the most talent and tracks where these graduates live after graduation.
The goal is to provide data-driven insights into talent availability and retention for companies,
startups, and policymakers.

Problem:
Germany faces a significant shortage of skilled workers, which is a major challenge for startups and
technology companies trying to grow. Businesses often lack clear data on where to find and recruit
top talent, as it's unknown whether graduates stay near their universities or move to other regions.
This study addresses this information gap by tracking the actual movement of skilled graduates,
helping companies make better strategic decisions about location and hiring.

Outcomes:
- Certain universities, particularly TU München, are talent powerhouses, producing the highest
number of graduates with both technical and entrepreneurial skills.
- Major metropolitan areas like Berlin, Munich, and Hamburg are very effective at retaining the
talent educated in their cities, with some Berlin universities keeping nearly 70% of their
entrepreneurial graduates locally.
- Germany's primary tech hubs—Berlin and Bavaria—are highly successful at not only retaining
their own graduates but also attracting a large number of skilled workers from other states.
- North Rhine-Westphalia (NRW) produces the largest volume of skilled graduates but has a lower
retention rate, effectively becoming a key talent supplier for Berlin and Bavaria.
- Cities and states often retain entrepreneurial talent at a higher rate than technical talent,
suggesting a potential gap in attractive job opportunities within the local tech industry.

Podcast:

About:
This study explores how digital technologies shape the entrepreneurial identities of women in the
tech industry. Using semi-structured interviews with 15 women entrepreneurs in Germany, the
research investigates the positive and negative impacts of digital tools on their professional
self-perception, mindset, and well-being.

Problem:
While digital technologies are often praised for making entrepreneurship more accessible, women
remain underrepresented in the field. Existing research has focused on structural barriers like
funding, but less is known about how the constant use of digital platforms directly influences a
woman's identity as an entrepreneur, including the unique pressures and opportunities it creates.

Outcomes:
- Digital technologies have a dual impact on women entrepreneurs: they provide empowering
flexibility but also create new pressures like constant connectivity and mental fatigue.
- Women use digital tools to balance motherhood and their careers ('digital motherhood'), but still
face biases and a lack of support structures in the digital space.
- Developing a 'digital mindset' is crucial, which is fostered through online access to role models
and safe, supportive digital communities for women.
- Constant online presence and the need to manage a digital professional brand contribute to stress
and potential burnout, negatively impacting mental well-being.

Podcast:

About:
This study analyzes how ten leading German companies translate the principles of Corporate
Digital Responsibility (CDR) into concrete business practices. Using a qualitative analysis of public
company data, the research examines the corporate governance structures, rules, and roles these
firms have established to manage their digital technologies ethically and responsibly.

Problem:
As businesses adopt new digital technologies, they create new ethical and societal challenges,
from AI bias to data privacy concerns. While the concept of Corporate Digital Responsibility (CDR)
provides a framework to address these issues, there is a significant gap between theoretical
principles and practical implementation, leaving companies without a clear roadmap for embedding
digital responsibility into their operations.

Outcomes:
- Companies are formalizing CDR with official C-level commitments and creating tailored
guidelines, values, and principles to steer their digital activities.
- Effective governance structures are emerging, including creating central points of contact,
establishing expert groups (e.g., Digital Ethics Panels), and ensuring accountability links directly to
senior management.
- Implementation involves creating official decision-making processes for ethical dilemmas,
continuously engaging stakeholders, and developing employee competence through awareness
campaigns and training.
- Leading firms are integrating CDR into risk management processes and performance reporting,
often as part of their broader sustainability efforts.
- The research derives seventeen practical learnings that serve as a model for how businesses can
structure their own CDR initiatives.

Podcast:

About:
This study details the design and initial evaluation of 'PharmAssistant,' a digital assistant created to
help pharmacists conduct medication reviews. Using a Design Science Research approach, the
researchers developed a prototype to streamline the time-consuming process of gathering patient
medication data prior to a consultation. The prototype's design was informed by interviews with
pharmacists and evaluated through focus groups with pharmacy students.

Problem:
Polypharmacy, the use of five or more medications, is a growing health concern, and medication
reviews are a key intervention to prevent drug-related problems. However, these reviews are not
widely practiced because they are extremely time-intensive for pharmacists, who often lack the time
and resources to manually collect comprehensive patient medication histories during consultations.

Outcomes:
- The 'PharmAssistant' prototype demonstrates a feasible way to automate patient data collection
before a medication review, potentially reducing the time burden on pharmacists.
- Features designed for older users, such as scanning medication barcodes (PZN numbers) and
using pre-defined answers, were identified as strengths that simplify data entry.
- Pharmacists expressed mixed opinions, acknowledging the tool's potential to save time but raising
concerns about usability for older patients and the importance of in-person interaction.
- Feedback from pharmacy students suggested key improvements, including adding lifestyle
questions, defining medical terms, ensuring robust data privacy, and integrating with electronic
health records.
- The research concludes that while digital assistants can support medication reviews, their success
depends on a user-centric design that is intuitive, empathetic, and accessible.

Podcast:

About:
This research develops a comprehensive taxonomy to structure and understand the applications of
Artificial Intelligence for Environmental Sustainability (AIfES). Based on an iterative analysis of
scientific literature and real-world examples, the authors created a multi-layer framework with nine
dimensions. This taxonomy serves as a tool for researchers and practitioners to holistically analyze,
classify, and understand AI systems designed for environmental purposes.

Problem:
While Artificial Intelligence (AI) is increasingly recognized as a key enabler for environmental
sustainability, the field lacks a systematic structure. Research and applications are often developed
for specific contexts, resulting in a fragmented landscape that makes it difficult for businesses and
researchers to get a comprehensive overview, compare solutions, or guide future development.
This absence of a common framework hinders the effective and responsible deployment of AI to
address pressing environmental challenges.

Outcomes:
- The study introduces a multi-layer taxonomy for classifying AI systems for Environmental
Sustainability (AIfES).
- This taxonomy is structured into three layers: 'Context' (the sustainability challenge being
addressed), 'AI Setup' (the technological foundation and data requirements), and 'Usage' (potential
risks and end-users).
- The framework contains nine distinct dimensions and 50 characteristics, enabling a detailed and
holistic analysis of AI applications.
- For businesses, the taxonomy offers practical guidance for designing, evaluating, and
communicating the sustainability value of their AI systems.
- The framework establishes a common language for technical, managerial, and sustainability
stakeholders to support the responsible development of impactful AI.

Podcast:

About:
This study conducts a systematic literature review of 57 academic papers to provide a structured
overview of how IT organizations can be designed for agility. The research identifies 20 distinct
agile design options and their associated performance benefits. These options are categorized into
four key dimensions: processes, structure, people & culture, and governance.

Problem:
In the digital age, traditional IT organizations often struggle to keep up with rapidly changing market
demands, hindering business innovation. While many companies recognize the need for greater
agility, the existing literature on how to redesign an IT organization is fragmented and lacks a clear,
consolidated framework. This makes it difficult for leaders to make informed, cost-effective
decisions when planning an agile transformation.

Outcomes:
- The study identifies 20 specific, actionable design options that IT organizations can implement to
become more agile, serving as a practical toolkit for leaders.
- These options are organized into four key areas: Processes (e.g., value stream orientation),
Structure (e.g., BizDevOps, product orientation), People & Culture (e.g., T-shaped skills), and
Governance (e.g., team autonomy, decentralized decisions).
- Each design option is explicitly linked to its expected performance benefits, such as increased
delivery speed, improved innovation, better alignment between business and IT, and greater
customer focus.
- The research emphasizes that a one-size-fits-all approach is ineffective; organizations must
carefully select and combine different design options tailored to their unique context.
- Key structural changes like adopting BizDevOps (integrating Business, Development, and
Operations) and shifting to a product-oriented model are shown to significantly improve end-to-end
responsibility and market responsiveness.

Podcast:

About:
This study introduces a new category of tool called regulatory support tools, designed to help
startups navigate complex regulations. The researchers developed and evaluated the 'Digital
Health Regulatory Navigator (EU)' using a Design Science Research approach to assist digital
health startups in understanding and strategically applying EU medical device regulations.

Problem:
Digital health startups, particularly in the European Union, face significant barriers to market entry
due to increasingly complex and strict medical device regulations. These early-stage companies
often lack the resources and legal expertise to navigate this environment, which can stifle
innovation and hinder the commercialization of beneficial technologies.

Outcomes:
- A new type of tool, the 'regulatory support tool', was conceptualized and designed to assist
startups in highly regulated industries.
- The 'Digital Health Regulatory Navigator (EU)' was developed as a practical tool to help digital
health startups understand EU medical device regulations and formulate a regulatory strategy.
- Evaluations with experts and entrepreneurs confirmed the tool's practical value, effectiveness, and
usability for its target audience.
- The study provides generalizable design principles that can be used to create similar support tools
for other regulated fields, such as those affected by the EU AI Act.
- Such tools can reduce compliance costs, help startups innovate more quickly, and strategically
leverage regulations for a smoother market entry.

Podcast:

About:
This study surveyed 314 recreational and competitive cyclists to understand what factors would
influence them to use virtual reality (VR) for indoor training. The research used an extended
Technology Acceptance Model (TAM) to identify the key drivers for adopting immersive training
technologies, using the popular platform Zwift as a hypothetical scenario.

Problem:
Current digital indoor cycling platforms are popular but offer a limited, screen-based experience.
While VR technology could make training more immersive and motivating, its potential in sports is
largely untested, creating uncertainty for businesses about whether cyclists would actually adopt it
and what features they would value.

Outcomes:
- The single most important factor for cyclists adopting VR is perceived enjoyment; making the
experience fun is more critical than its utility for performance improvement.
- Perceived usefulness is a significant predictor of adoption, but less influential than enjoyment.
- Surprisingly, how easy the technology is to use does not significantly impact a cyclist's intention to
use it.
- Social factors, such as recommendations from peers and trainers, strongly influence how useful
and enjoyable cyclists perceive VR to be.
- An individual's general openness to new technology is a key factor in predicting positive reception.

Podcast:

About:
This study investigates how to design effective systems for monitoring organizational change
projects. Using an action design research approach with two German manufacturing companies,
the researchers developed and evaluated a prototype system consisting of surveys and an
interactive dashboard. The goal was to derive generalized requirements and design principles to
guide practitioners and academics in creating similar tools.

Problem:
Organizations constantly undergo change to remain competitive, but managing these initiatives is
difficult. While IT systems can monitor change projects to improve transparency and
decision-making, there is very little research or practical guidance on how to design these systems
effectively. This knowledge gap hinders the development of tools that can successfully support
complex change management processes.

Outcomes:
- The study produced a prototype monitoring system using surveys and a Power BI dashboard to
track key metrics like change readiness, acceptance, and implementation.
- Key design challenges were identified, including balancing the need for detailed insights with user
effort, accommodating different change management styles, and managing expectations about
data-driven analysis.
- Three core requirements for change monitoring systems were proposed: they must be adaptable
to different users and projects, provide information tailored to various stakeholder needs, and
minimize the effort required to use them.
- The research provides eight specific design principles for building these systems, focusing on
aspects like using common data formats, ensuring modularity, involving stakeholders throughout
the process, and communicating a clear vision for the tool.

Podcast:

About:
This study investigates the integration of Generative AI (GenAI) into knowledge work, focusing on
the resulting conflicts between benefits and risks. Using socio-technical systems theory, the
research identifies key tensions through a systematic literature review and interviews with 18
knowledge workers. The paper proposes solutions to help organizations navigate these challenges
for responsible and effective AI adoption.

Problem:
While Generative AI offers significant opportunities to increase productivity and efficiency in
knowledge work, its adoption introduces complex challenges. Organizations face inherent conflicts
between leveraging AI for speed and ensuring the reliability of outputs, protecting sensitive data,
and maintaining ethical standards. This creates uncertainty and risk, highlighting the need for clear
strategies to balance innovation with responsible governance.

Outcomes:
- Productivity vs. Reflection: GenAI boosts efficiency but can lead to over-reliance and reduced
critical thinking, potentially compromising work quality.
- Availability vs. Reliability: Constant access to GenAI does not guarantee accurate information,
increasing the business risk of acting on misinformation.
- Convenience vs. Data Protection: The ease of using GenAI for daily tasks creates a significant risk
of exposing confidential or sensitive company data to external systems.
- Efficiency vs. Traceability: AI generates content quickly, but the lack of verifiable sources makes it
difficult to trust and use in professional settings where accuracy is critical.
- Regulatory Freedom vs. Ambivalence: The widespread accessibility of GenAI clashes with a lack
of clear regulations, creating legal and ethical uncertainties for companies regarding its responsible
use.

Podcast:

About:
This study examines why people choose to trust or ignore AI advice when identifying deepfake
videos. Through an online experiment, researchers presented participants with videos and an AI
detection tool's recommendation, analyzing how factors like trust, AI literacy, and algorithm
aversion influenced their final judgments.

Problem:
The rise of highly realistic deepfakes makes it increasingly difficult to distinguish real from
manipulated content, posing significant threats like the spread of misinformation and reputational
damage. While AI detection tools are available, little is known about the psychological factors that
drive user reliance on these systems in high-stakes situations.

Outcomes:
- Users were far more likely to follow an AI's advice when it confirmed a video was genuine, but
they often ignored its advice when it flagged a video as a deepfake.
- This tendency to accept 'genuine' ratings from AI occurred regardless of the user's level of trust, AI
literacy, or general aversion to algorithms.
- Reliance on AI advice to confirm authenticity was driven by different combinations of user traits,
including having low trust, high AI literacy, or high aversion to algorithms.
- The findings suggest users may treat AI tools as confirmatory aids for authenticity rather than as
reliable detectors of manipulation, creating a risk of 'blindly following' AI recommendations.

Podcast:

About:
This study investigates how to encourage more deliberate, analytical thinking when people use
Generative AI (GenAI) for decision-making. Through an experiment with 130 participants,
researchers tested a "sequential" interaction design where users made an initial decision before
receiving GenAI assistance. This approach was compared against receiving simultaneous AI help
or having no AI help at all.

Problem:
When using Generative AI for decision support, people often rely on quick, intuitive judgments
(System 1 thinking) instead of careful, analytical thought (System 2). This over-reliance on "gut
feelings" can lead to suboptimal decisions and fails to leverage the full potential of AI assistance.
The study addresses the challenge of designing AI interactions that encourage users to think more
reflectively and critically.

Outcomes:
- Requiring users to make an initial decision before receiving GenAI help significantly improves
decision-making performance compared to getting AI help simultaneously or having no AI at all.
- This sequential interaction method encourages users to think more critically and engage more
deeply with the AI, as shown by significantly higher use of prompts.
- Making an initial, unassisted choice helps users recognize conflicts between their intuitive answer
and the AI's suggestion, prompting a shift to more deliberate, analytical thinking.
- This 'nudge' leads to better final decisions without restricting user autonomy, offering a practical
design principle for more effective human-AI collaboration systems.

Podcast:

About:
This study analyzes social biases within modern chat-based AI models, comparing their
performance in both English and Spanish. The researchers developed a systematic method to test
three leading AI models using specialized datasets. The evaluation focused on the models' ability to
identify and refuse to answer biased prompts, as well as the fairness of their responses when they
do answer.

Problem:
As businesses increasingly rely on AI for critical decisions in areas like human resources and
customer service, there is a significant risk that these systems could perpetuate harmful social
stereotypes, leading to unfair outcomes. Most research on AI bias has focused exclusively on
English, creating a major knowledge gap for other globally important languages like Spanish and
posing risks for international companies deploying these technologies.

Outcomes:
- AI models surprisingly provide fairer answers in Spanish than in English.
- However, the models are less effective at identifying and refusing to engage with biased questions
when prompted in Spanish.
- AI fairness improves significantly when questions are direct and unambiguous, as opposed to
indirect or vague.
- When provided with enough clear context, all tested models can overcome initial stereotypes to
give highly fair answers.
- A model's general accuracy on neutral tasks does not guarantee its fairness, indicating a potential
trade-off between performance and ethical behavior.

Podcast:

About:
This study investigates how industrial product companies are adopting Generative AI (GenAI) by
identifying key obstacles and practical solutions. Based on expert interviews with industry leaders
and technology providers, the research analyzes challenges across technology, organization, and
environment. The findings offer guidance for companies to move beyond experimentation and
create meaningful value with GenAI.

Problem:
While GenAI is transforming many industries, its adoption in industrial product companies is
particularly difficult. These firms often lack deep digital expertise and must integrate new technology
with complex legacy systems involving hardware, software, and services. This creates a significant
gap between the hype surrounding GenAI and the practical challenges of realizing its full impact.

Outcomes:
- Overcome technical issues like AI 'hallucinations' and inconsistent results by grounding models
with internal company data (RAG) and implementing standardized testing and prompt logging.
- Address organizational challenges by shifting from complex ROI calculations to simpler KPIs like
user adoption and time saved, and by managing hype through realistic employee training.
- Mitigate the 'make vs. buy' dilemma by compelling tech vendors to share their development
roadmaps, enabling better strategic alignment and avoiding investment in quickly outdated in-house
solutions.
- Navigate external risks like vendor lock-in and regulatory uncertainty by building model-agnostic
architectures to easily switch between AI providers and establishing clear compliance frameworks.

Podcast:

About:
This study investigates how Generative AI (GenAI) tools like ChatGPT and GitHub Copilot affect
knowledge management and team performance in software development. Through a survey of 80
software developers, the research examines the relationship between GenAI usage, the processes
of knowledge transfer and application, and overall team effectiveness.

Problem:
While the productivity benefits of GenAI for individual tasks are becoming clear, its broader impact
on team-level knowledge processes and performance is not well understood. Without a clear
understanding, businesses risk negative consequences like the erosion of human expertise or
reduced knowledge retention when integrating these powerful tools.

Outcomes:
- Using GenAI tools has a significant positive impact on both knowledge transfer (sharing
information) and knowledge application (using information to solve problems) within software
teams.
- The use of GenAI directly enhances overall team performance, primarily by improving how teams
apply knowledge.
- Simply transferring knowledge using GenAI does not significantly improve team performance on
its own; the knowledge must be actively applied to be effective.
- GenAI acts as a 'cognitive extension' for teams, helping them work more efficiently, but
businesses must manage its integration thoughtfully to avoid over-reliance and maintain valuable
human-to-human collaboration.

Podcast:

About:
This study replicates and validates a decade-old set of metrics for analyzing user activity in digital
collaboration tools. By applying measures like activity and productivity to a modern platform, the
research confirms their continued relevance for understanding team collaboration. The study also
introduces and tests a new method, Collaborative Work Codes (CWC), to provide more nuanced
insights into how different types of teams (e.g., projects, departments) use these digital
workspaces.

Problem:
With the rise of remote and hybrid work, businesses increasingly rely on digital collaboration
software, which generates vast amounts of user data. However, these platforms typically lack
built-in analytical tools, leaving managers without clear insights into how their teams are working,
collaborating, or engaging with the tools, making it difficult to optimize workflows and support
productivity.

Outcomes:
- A decade-old set of metrics for measuring activity, productivity, and cooperation in digital
workspaces remains valid and effective on modern collaboration platforms.
- Different types of workgroups (projects, departments, and training courses) show distinct and
predictable patterns of activity, which can be used to identify and profile them.
- A new method using Collaborative Work Codes (CWC) offers a more detailed understanding of
user actions (e.g., retrieving information vs. creating content), enhancing workspace analysis.
- The study confirmed the Pareto principle (the '80-20 rule') in project and departmental
workspaces, where a small number of users generate the majority of the activity.
- A practical framework, called a 'morphological box', was developed to help managers classify their
digital workspaces based on attributes like team size and activity patterns, enabling better
management.

Podcast:

About:
This study investigates how the layout of a website, known as a Digital Choice Environment (DCE),
affects customer choices. Using an online experiment with 421 participants in a fictional
e-commerce store, the researchers tested how the presence and placement of website elements,
like a chatbot, influence purchasing decisions and interact with marketing nudges such as
'Bestseller' tags. The goal was to understand the subtle but significant role of overall website
configuration on user behavior.

Problem:
Businesses frequently add elements like chatbots or promotional tags to their websites to guide
customer choices, often without considering the existing layout. It's largely unknown how the overall
website design interacts with these new elements, potentially leading to ineffective or
counterproductive results. This research addresses the gap in understanding how the complete
digital environment shapes the effectiveness of individual design choices and interventions.

Outcomes:
- The mere presence of a website component, like a chatbot, significantly alters customer
purchasing decisions. In the experiment, adding a chatbot approximately doubled the likelihood of a
specific product being chosen.
- The placement of website components is critical. For example, a chatbot placed on the left side of
the screen led to different product choices than one placed on the right.
- A chatbot and a marketing nudge (a 'Bestseller' tag) influenced choices independently. The
chatbot did not weaken the effectiveness of the nudge, suggesting multiple design elements can
work in parallel without interference.
- Businesses must consider the website's entire layout as a cohesive system, as even small
changes to one component can have unexpected effects on user behavior and the success of
marketing efforts.

Podcast:

About:
This study analyzes existing research to create a framework for understanding the various
Generative AI (GenAI) tools used in business innovation. Through a structured literature review, the
paper proposes a classification system based on socio-technical theory. The goal is to bring clarity
to the crowded market of GenAI applications and help managers integrate them effectively.

Problem:
The rapid proliferation of Generative AI (GenAI) tools has created a confusing landscape for
businesses. Innovation managers lack a clear understanding of the different types of tools
available, making it difficult to select, compare, and effectively integrate them into their innovation
processes.

Outcomes:
- GenAI tools can be categorized into four key dimensions to guide selection and implementation.
- People: Tools differ based on whether they are for individuals or teams and how users interact
with the AI (e.g., as a co-creator or delegator).
- Organization: Classification should consider how a tool fits within a company, including data
security, ethics, and industry-specific requirements.
- Task: Tools support different innovation stages, such as idea generation (divergent thinking) or
evaluation (convergent thinking).
- Technology: The technical capabilities vary, including the level of automation versus human
augmentation, transparency, and compatibility with other systems.

Podcast:

About:
This study investigates why and how knowledge workers use "digital detox" to manage workplace
stress caused by technology (technostress). Based on 16 semi-structured interviews with
professionals, the research identifies the primary motivators and necessary conditions for
successfully disconnecting from work-related digital tools. The paper explores digital detox as a
coping strategy that enables psychological detachment in the modern workplace.

Problem:
The constant connectivity and information overload from digital technologies in the workplace are
leading to increased employee stress, burnout, and reduced performance, a phenomenon known
as technostress. Organizations lack a clear understanding of why employees seek to disconnect
and what support they need to practice digital detox effectively without compromising flexibility or
productivity.

Outcomes:
- The primary motivators for digital detox are to improve both well-being (by mentally switching off
from work) and job performance (by increasing focus and efficiency).
- Digital detox is a direct response to specific stressors like 'techno-overload' (too much information
and pace) and 'techno-invasion' (work encroaching on personal time).
- Successful digital detox requires a combination of individual responsibility (e.g., discipline,
self-control) and organizational responsibility (e.g., supportive policies, leadership buy-in, and a
culture that respects downtime).
- Both individuals and their organizations must work together to enable effective digital detox, as
individual efforts can be undermined by organizational expectations of constant availability.
- Flexibility is crucial, allowing employees to choose detox strategies (from turning off notifications
to scheduling focus time) that suit their specific roles and work demands.

Podcast:

About:
This study explores how responsibility evolves in human-AI collaboration, particularly addressing
the 'responsibility gap' where accountability is unclear. Using an affective agency perspective, the
research examines how human emotions and relationships with technology shape the assignment
of responsibility. The methodology involved qualitative interviews with experts across various
professional sectors to understand real-world practices.

Problem:
When AI systems influence or make autonomous decisions, it becomes difficult to assign blame for
errors, creating a 'responsibility gap.' This ambiguity poses significant ethical and legal challenges
in critical fields, as traditional accountability models that rely on a single human decision-maker are
no longer adequate. The study addresses the urgent need to understand how responsibility can be
clearly defined and maintained in collaborative human-AI environments.

Outcomes:
- Instead of diminishing human accountability, the use of AI often intensifies it, compelling users to
be more critical and vigilant in validating the system's outputs.
- Professionals view AI as a supportive tool or 'sparring partner,' not a replacement for human
judgment; the final decision and ultimate responsibility remain firmly with the human user.
- The inherent uncertainty of AI systems encourages a cautious and engaged approach, where
emotions like trust and skepticism actively shape how responsibility is assumed.
- A model of shared responsibility is emerging, where developers are accountable for transparency
and ethical design, while users are responsible for the contextual application of AI tools.
- The study highlights a clear need for stronger regulations and organizational guidelines to formally
define roles and liabilities in human-AI collaboration.

Podcast:

About:
This study investigates the unintended psychological effects of Privacy-by-Default (PBD) settings in
mobile applications. Using information boundary theory, the research explores how automatically
enforced privacy controls can paradoxically reduce a user's sense of autonomy. The study
proposes an online field experiment to test how PBD influences a user's intention to disclose
personal information.

Problem:
While Privacy-by-Default is widely considered a best practice for protecting user data, it may create
a conflict for users. By removing active choice, these default settings can undermine an individual's
sense of control and autonomy over their personal information, potentially leading to unintended
behaviors and challenging the design of effective privacy systems.

Outcomes:
- Privacy-by-Default (PBD) can create an 'autonomy-disclosure paradox': while it protects users, it
may also reduce their sense of control.
- A reduced sense of autonomy can negatively impact a user's decision-making regarding
information sharing.
- The context (e.g., professional vs. private app usage) significantly influences how users react to
default privacy settings.
- The research challenges the straightforward assumption that PBD is always beneficial for users,
highlighting that it can diminish their sense of agency.
- App designers and policymakers should aim for a balance between protective default settings and
meaningful user empowerment to avoid negative psychological effects.

Podcast:

About:
This study investigates how users respond when they perceive their personal information has been
used for a secondary purpose by an external party (External Secondary Use or ESU), which they
consider a privacy violation. Using Qualitative Comparative Analysis (QCA) and attribution theory,
the research identifies specific combinations of factors—such as the user's emotions, expectations,
and sense of control—that lead them to either restrict data sharing or stop using a digital service
altogether.

Problem:
Digital services frequently collect user data for one purpose and then share it with third parties for
other uses, like targeted advertising. This practice, known as External Secondary Use (ESU), is a
growing privacy concern that can erode user trust. When users feel their privacy has been violated,
they may take actions that negatively impact the business, such as discontinuing the service or
restricting data collection, but it's often unclear what specific factors trigger these different negative
responses.

Outcomes:
- Different emotional responses lead to distinct user actions: anxiety often results in users restricting
their information sharing while continuing to use the service, whereas anger is a strong predictor of
users discontinuing the service entirely.
- Users are more likely to stop using a service if the privacy violation was unexpected (e.g.,
unsolicited phone calls). If they anticipated such data use, they are more likely to restrict their data
but remain a customer.
- A user's belief in their ability to take protective action (self-efficacy) plays a key role. When angry
users feel empowered to act, they are more inclined to leave the service.
- The type of violation matters. Unsolicited direct contact is perceived as more severe and is more
likely to cause users to leave, compared to receiving targeted web ads.
- Businesses can mitigate negative outcomes by being transparent about data sharing practices to
manage user expectations and by addressing the specific causes of user anger to improve
customer retention.

Podcast:

About:
This study explores the 'venture client' model, originally from the private sector, as a method for
public institutions to accelerate the adoption of new technologies. Through a case study of the first
dedicated public sector venture client unit, the research uses interviews with government
technology startups (GovTechs) to assess the model's potential to speed up procurement and
foster innovation. The goal is to understand if this model is a viable alternative to traditional, often
slow, procurement processes.

Problem:
Public organizations face significant pressure to innovate and adopt modern technology to meet
citizen demands, but are often hindered by slow, bureaucratic, and risk-averse procurement
processes. While various innovation strategies exist, there is a lack of academic research and
empirical evidence on how the venture client model—collaborating with startups without direct
investment—can be effectively applied in the unique context of the public sector. This study aims to
fill that gap by providing an initial analysis of the model's real-world application and benefits.

Outcomes:
- The venture client model dramatically accelerates technology procurement, reducing timelines
from over two years to just three to six months.
- This approach provides significant benefits to government-focused startups (GovTechs) by
shortening their time-to-market and helping them scale solutions tailored to public administration
needs.
- Interviews with GovTechs identified key success factors for working with the public sector,
including having a reference customer, access to accelerated procurement, and dedicated
'sandboxing' budgets for pilot projects.
- The research is ongoing, and it is not yet clear how the model must be adapted for specific public
sector use-cases to deliver maximum value.

Podcast:

About:
This study analyzes 48 real-world examples of circular economies to identify common patterns of
collaboration between different organizations. Using e³-value modeling, the research identifies and
categorizes eight recurring business models, or "constellations". These models provide practical
blueprints for organizations aiming to transition towards a circular economy.

Problem:
While the circular economy offers a promising path to sustainability, businesses often struggle to
implement it effectively because it requires complex collaboration with new types of partners. There
is a lack of clear, structured understanding of how different actors (like producers, consumers, and
recyclers) should interact and exchange value, which hinders the successful adoption of circular
practices.

Outcomes:
- The research identifies eight distinct collaboration models ('constellations') for circular economies,
grouped into three overarching categories.
- The first category, 'Circularity-driven Innovation', includes models led by producers, service
providers, or regulators to create more sustainable products and processes.
- The second, 'Resource Efficiency Optimization', focuses on models for sharing and redistributing
underutilized resources, either through an intermediary or peer-to-peer platforms.
- The third, 'End-of-Life Product & Material Recovery', outlines models for collecting, processing,
and reintegrating used products back into the economy, led by specialized 'scavenger' or
'decomposer' actors.
- These models serve as strategic blueprints, helping companies identify necessary partners and
define how value is created and exchanged in a circular system.

Podcast:

About:
This study provides a preliminary analysis of the ethical challenges in smart city development
through a systematic literature review. The research method involves identifying, categorizing, and
synthesizing academic papers that discuss the value conflicts faced by various stakeholders. The
goal is to understand the recurring ethical tensions, such as security versus privacy, to inform more
ethically-grounded urban planning.

Problem:
While smart cities promise to improve urban life with advanced technologies, their development
often neglects complex ethical dilemmas. This creates a gap where technological efficiency is
prioritized over human values, leading to potential issues with surveillance, social equity, and
democratic participation. The lack of a clear ethical framework to guide these projects means that
the needs and rights of diverse city inhabitants can be overlooked.

Outcomes:
- The most frequently discussed ethical conflict in smart city literature is the tension between
increasing public security through surveillance and protecting citizens' right to privacy.
- The pursuit of technological efficiency often leads to social exclusion, as systems designed for
optimization may reinforce existing inequalities and fail to serve all community members equitably.
- Many smart city projects are driven by a 'solutionist' mindset that prioritizes rapid innovation over
ethical oversight and public deliberation, potentially undermining democratic legitimacy.
- Ethical tensions are not isolated incidents but are often structurally embedded in the design and
governance of smart city systems, highlighting a systemic challenge.
- There is a broad consensus on the need for transparent, multi-stakeholder processes that make
value conflicts explicit and allow for public debate and accountability.

Podcast:

About:
This study develops a decision-making framework, called a taxonomy, to help university educators
systematically evaluate if Virtual Reality (VR) is appropriate for their specific teaching content. The
taxonomy was created by combining established educational theories with insights from expert
interviews and a literature review. The resulting framework helps classify reasons for and against
using VR across five dimensions, guiding educators in their technology adoption decisions.

Problem:
Despite growing interest in using Virtual Reality (VR) in higher education, many lecturers lack a
structured method to determine if the technology is actually suitable for their specific courses. This
creates a significant gap, as decisions to adopt VR are often made without a clear understanding of
its potential benefits or limitations for a given subject. Consequently, educators, especially those
without prior VR experience, struggle to make informed choices.

Outcomes:
- The research produced a comprehensive taxonomy to guide educators in assessing VR's
suitability for specific learning content.
- The framework is organized into five key dimensions: learning objectives, learning activities,
learning assessment, social influence, and hedonic motivation.
- It identifies specific scenarios where VR is beneficial, such as simulating dangerous or
inaccessible environments, enabling practice of complex skills, and creating immersive, focused
learning experiences.
- The taxonomy also outlines contexts where VR is less suitable, such as for content requiring
nuanced social learning, extensive reading, or collaborative writing tasks.
- This tool acts as a practical checklist to help educators make a balanced, initial assessment
before committing significant resources to developing VR-based teaching materials.

Podcast:

About:
This study presents and evaluates a three-phase Decision Support System (DSS) designed to
automatically analyze corporate annual reports (10-Ks). The system uses Natural Language
Processing (NLP) to extract forward-looking statements, machine learning to predict future financial
metric growth, and Generative AI to summarize the findings for users. The methodology was tested
on 10-K reports from S&P; 500 companies filed between 2015 and 2022.

Problem:
Investors and analysts rely on lengthy and complex 10-K reports to make financial decisions, but
manually extracting key predictive information from the narrative sections is inefficient and difficult.
Existing automated tools often lack transparency or fail to provide specific, actionable forecasts tied
to financial metrics. This research addresses the need for an automated system that can transform
unstructured text into transparent, metric-specific performance predictions.

Outcomes:
- The system extracted forward-looking statements related to specific financial metrics with 94%
accuracy.
- A Random Forest machine learning model outperformed a more complex FinBERT model in
predicting future metric growth, indicating that simpler models can be more effective for this task.
- Using textual forward-looking statements significantly improved prediction accuracy compared to a
baseline model that only used past financial data.
- AI-generated summaries of the financial outlook were rated highly for factual consistency and
readability (average score of 3.69 out of 4), making the system's output more transparent and
useful for decision-makers.

Podcast:

About:
This study uses Multi-Criteria Decision Analysis (MCDA) to evaluate and compare four distinct
approaches for governing algorithmic management systems in the workplace. Based on a
structured questionnaire with 27 experts from academia, industry, and government, the research
assesses principle-based, rule-based, risk-based, and auditing-based models against criteria like
effectiveness, feasibility, and adaptability.

Problem:
Companies increasingly use algorithms to manage workers, which creates efficiency but also raises
concerns about fairness, transparency, and employee well-being. A significant gap exists in
understanding how to best govern these systems, as there is little empirical evidence comparing
different strategic approaches. This study addresses the lack of a clear, evidence-based framework
for organizations to choose the most effective and responsible governance model.

Outcomes:
- Experts overwhelmingly prefer a hybrid, risk-based approach for governing algorithmic
management systems, as it was ranked highest in effectiveness, adaptability, and stakeholder
acceptability.
- This hybrid model successfully balances the flexibility of high-level principles with the clarity of
specific rules, tailoring governance intensity to the level of potential harm.
- Purely rule-based systems were deemed too rigid and slow to adapt to new technology, while
purely principle-based approaches were considered too abstract and difficult to enforce.
- The primary business recommendation is to abandon 'one-size-fits-all' policies. Instead,
companies should classify their algorithmic tools by risk level and apply stricter controls (e.g.,
human oversight, bias audits) only to high-risk applications like automated hiring or firing.

Podcast:

About:
This study analyzes 36 contemporary mobile habit-formation apps to understand how they support
lifestyle changes across multiple, interconnected domains. Using content analysis, researchers
categorized 585 distinct behavior recommendations into 20 overarching categories. The goal was
to identify key patterns in app design to provide insights for both users and developers on selecting
and creating more effective, holistic habit-tracking tools.

Problem:
Successfully adopting one positive habit can create a ripple effect, leading to improvements in other
areas of life, such as exercise improving diet and sleep. However, it is unclear if and how current
mobile habit-tracking apps are designed to leverage these interconnected behaviors to support
comprehensive lifestyle changes. This research addresses the gap by examining which behaviors
are recommended by apps and how they are linked together.

Outcomes:
- Physical Exercise is the most dominant and frequently recommended behavior in habit apps,
serving as a central anchor that is strongly linked to Nutrition, Leisure Activities, and Sleep.
- Most habit apps encourage a multidomain approach, offering suggestions across an average of
13 different lifestyle categories, indicating a trend toward holistic well-being.
- Apps that cover a broader range of lifestyle domains also tend to offer more advanced
habit-formation features and functionalities.
- User satisfaction is not guaranteed by functional diversity; simply offering more features or a wider
variety of habit suggestions does not directly correlate with higher app store ratings.
- Certain goals, like weight management, are often treated in isolation within apps rather than being
integrated into a broader, interconnected lifestyle approach.

Podcast:

About:
This study presents a framework for safely integrating Artificial Intelligence (AI) agents into data
trusts, which are organizations that manage data on behalf of others. It proposes four key design
principles derived from fiduciary duties and AI ethics to guide the development of trustworthy
AI-powered governance systems. The goal is to balance the efficiency gains from AI automation
with the need for responsible and ethical data stewardship.

Problem:
As data trusts emerge to manage sensitive information, there's a push to use AI for greater
efficiency. However, using AI raises significant risks, such as algorithmic bias, lack of transparency
(the "black box" problem), and potential conflicts of interest, which can erode the trust essential for
these institutions to function. This paper addresses the lack of a clear guide on how to deploy AI in
a data trust without compromising the fundamental duties of loyalty and care owed to data owners.

Outcomes:
- Ensure AI acts in the best interest of data owners: AI systems must be designed with 'fiduciary
alignment,' prioritizing the beneficiaries' interests over commercial or other external objectives.
- Maintain accountability with full traceability: Since AI agents cannot be held legally responsible, all
their decisions must be logged and auditable, ensuring that human operators can be held
accountable.
- Make AI operations transparent and explainable: The system should provide clear explanations
for AI-driven decisions and offer stakeholders visibility through dashboards to build and maintain
institutional trust.
- Preserve human autonomy and control: Data governance must allow for dynamic consent where
data owners can continuously oversee and adjust permissions, with human review required for
critical decisions and conflict resolution.

Podcast:

About:
This study empirically investigates how the use of Generative AI (GenAI) by management impacts
the social dynamics of business and IT collaboration. Using a literature review, an expert survey of
61 executives, and statistical modeling, the research analyzes GenAI's effect on communication,
shared knowledge, and mutual understanding between business and IT departments.

Problem:
While companies are rapidly adopting GenAI, its effect on the crucial human side of business-IT
alignment—how well teams communicate, understand each other, and collaborate—remains
largely unexplored. Poor alignment between business goals and IT capabilities can lead to
inefficient investments and missed opportunities, a problem GenAI could either solve or worsen.

Outcomes:
- GenAI significantly improves formal business-IT collaboration by enhancing structured knowledge
sharing and helping to develop a common language between departments.
- The technology helps bridge knowledge gaps, increasing business leaders' understanding of IT
concepts and IT leaders' grasp of business strategy.
- GenAI does not have a significant impact on informal, relationship-based interactions;
trust-building and networking still require human-driven leadership and engagement.
- To maximize value, management should strategically use GenAI to streamline formal
communication and knowledge sharing, while actively fostering a culture of interpersonal
collaboration.

Podcast:

About:
This study explores how AI-powered Personal Digital Assistants (PDAs), like chatbots, can improve
the way companies share knowledge about their business processes. Using qualitative interviews
with professionals from various industries, the research identifies nine key benefits (value
propositions) that PDAs offer for transferring this critical knowledge.

Problem:
In modern businesses, crucial knowledge about how work gets done is often buried in complex
documents, making it difficult for employees to access and use efficiently. Traditional methods for
sharing this 'process knowledge' are often slow and ineffective, leading to errors, inefficiencies, and
missed opportunities for improvement.

Outcomes:
- PDAs make process knowledge more accessible by automating routine tasks and allowing
employees to find information and documentation much faster.
- They improve understanding by educating users on processes, accelerating the onboarding of
new employees, and providing context-aware analysis that integrates with other business systems.
- PDAs offer direct guidance by providing step-by-step advice, suggesting process optimizations,
enforcing standardization, and aiding in data-driven decision-making.

Podcast:

About:
This study explores how to design Augmented Reality (AR) interfaces on sport glasses to help
endurance runners achieve a state of "flow," which is an optimal mental state for peak performance.
Using a Design Science Research approach, the authors created and tested an AR display
prototype with nine runners over two iterative design cycles. The research aimed to derive practical
design principles based on real-world user feedback.

Problem:
For endurance runners, achieving and maintaining a state of flow is challenging due to the dynamic
outdoor environment and the need for constant, non-distracting performance feedback. While AR
technology offers a potential solution by overlaying data into a user's field of view, there is a lack of
clear design guidelines on how to create AR interfaces that effectively support flow in a physically
demanding activity like running without causing distractions.

Outcomes:
- Provide easily interpretable, non-numeric feedback: Use visual cues like colors, shapes, and
animations instead of just numbers to help runners understand their performance at a glance and
reduce cognitive load.
- Ensure the interface is adaptive and unobtrusive: The AR display should minimize distractions,
adapt to different lighting conditions, and offer varying levels of feedback to suit both novice and
experienced runners.
- Allow for user customization: Runners should be able to choose which performance metrics they
see (e.g., heart rate, pace, distance) to align the feedback with their personal training goals.
- Maintain engagement with subtle novelty: To keep users interested over time, the interface should
incorporate dynamic but subtle design elements that are engaging without being distracting or
overwhelming.

Podcast:

About:
This study investigates how the widespread adoption of Generative AI (GenAI) tools impacts
external developer contributions to digital platforms. By analyzing the open-source activity of four
major e-commerce platforms on GitHub, the research measures changes in developer activity,
contributor diversity, and the balance between internal and external development following the
introduction of tools like ChatGPT.

Problem:
Digital platforms rely on external developers, or 'complementors', to create value and innovate. The
rapid rise of GenAI significantly boosts developer productivity, especially for those with less
experience, but it's unclear how this technological shift affects the volume and nature of external
contributions. This research addresses the gap in understanding GenAI's impact on platform
ecosystems and how platform owners should adapt.

Outcomes:
- There was a significant shift toward external value creation, with the ratio of contributions from
external developers increasing relative to internal development after GenAI's introduction.
- GenAI increased the diversity of contributors, leading to a positive shift in contributions from the
most active (top 1%) and moderately active (next 9%) developers across most platforms.
- The overall volume of external contributions showed a complex, heterogeneous effect, with initial
increases followed by a decrease after the widespread adoption of ChatGPT.

Podcast:

About:
This study investigates why people are reluctant to use algorithms, a phenomenon known as
'algorithm aversion'. Through an online experiment with 280 participants, the researchers tested
whether giving users control to adjust an algorithm's predictions or making the algorithm's
decision-making process transparent could increase user trust and adoption. The study aimed to
understand the separate and combined effects of adjustability and transparency on user behavior.

Problem:
Despite often outperforming humans, machine learning (ML) models are frequently underutilized
because people mistrust them. This 'algorithm aversion' is a significant barrier to adopting valuable
AI tools in business and other domains. While prior research showed that letting users adjust AI
outputs helps build trust, it was unclear if making the AI's logic transparent would be an equally or
more effective strategy.

Outcomes:
- Giving users control to adjust an algorithm's predictions is a highly effective way to reduce their
reluctance to use it.
- Users who could modify an algorithm's output were significantly more likely to rely on it, which also
led to better overall performance on the prediction task.
- In contrast, simply making the algorithm's decision process transparent ('opening the black box')
had a surprisingly small and statistically insignificant impact on user adoption.
- The study concludes that providing a tangible sense of control is more crucial for overcoming
algorithm aversion than just explaining how the model works.

Podcast:

About:
This study develops a comprehensive classification system, or taxonomy, for Embodied Generative
AI—intelligent systems that operate within physical bodies like robots. Based on a systematic
review of academic literature and an analysis of 40 real-world examples, the paper creates a
structured framework to categorize and understand these emerging technologies. The resulting
taxonomy provides a foundation for future research and offers a practical tool for analyzing and
optimizing applications.

Problem:
As Generative AI (GenAI) moves from purely digital tasks to controlling physical systems, a new
class of 'Embodied GenAI' is emerging in fields like robotics and automation. However, this is a new
and diverse field, and there has been no standardized way to classify or evaluate the different
features of these systems. This lack of a systematic framework makes it difficult for businesses and
researchers to compare products, guide development, and make informed decisions about
adopting this new technology.

Outcomes:
- The study created a detailed taxonomy for classifying Embodied GenAI systems.
- This framework is organized into three main categories: Embodiment (the physical form),
Intelligence (the AI's cognitive capabilities), and System (the integration of both).
- It identifies 16 key dimensions (e.g., appearance, autonomy, collaboration) and 50 specific
characteristics to describe and compare different systems.
- Analysis of 40 current systems revealed key trends, such as a majority (60%) having human-like
designs and a strong focus on fully autonomous operation (63%).
- The taxonomy provides a practical tool for businesses to assess existing technologies and a
reference for designing new applications in service robotics and industrial automation.

Podcast:

About:
This study synthesises existing research to provide an integrated framework for understanding
digital innovation. Through a structured literature review, the paper identifies five key 'catalysts' that
can either accelerate or hinder innovation, along with the tensions and trade-offs inherent in each.

Problem:
While digital innovation is crucial for competitive advantage, managers and scholars often lack a
holistic view of the interconnected factors that drive it. Research tends to focus on individual
phenomena like platform ecosystems or product design in isolation, making it difficult for
businesses to strategically manage the complex interplay of opportunities and risks.

Outcomes:
- The paper identifies five primary catalysts for digital innovation: Data Objects, Layered Modular
Architecture, Product Design, IT and Organisational Alignment, and Platform Ecosystems.
- Each catalyst introduces both positive stimuli (e.g., data monetization, rapid product updates,
network effects) and negative tensions (e.g., privacy risks, user confusion, market lock-in).
- Data monetization creates new revenue but raises privacy concerns, while flexible modular
architectures can lead to fragmentation if not governed properly.
- Innovative product designs can redefine value but risk confusing users if they are too complex or
ignore existing behaviors.
- Successfully managing digital innovation requires balancing these trade-offs, aligning technology
with organizational culture, and understanding the dynamics of the broader ecosystem.

Podcast:

About:
This study analyzed over 37,000 user reviews from 22 mobile health apps designed for
cardiovascular care. Using a data analysis technique called topic modeling, the research identified
key patterns in what users value and discuss, revealing the most important features and capabilities
for managing heart conditions.

Problem:
Cardiovascular disease is a leading cause of death, and mobile apps offer a promising way for
patients to monitor their health. However, for these apps to be effective, they must meet patient
needs. This study addresses the gap between technology availability and user adoption by
analyzing real-world feedback to understand what makes these apps truly useful for patients.

Outcomes:
- Data Management & Sharing: Users prioritize the ability to easily manage, document, and share
their health data (e.g., via PDF reports) with clinicians.
- Monitoring & Tracking: Core functionality for real-time measurement and long-term monitoring of
vital signs (like heart rate) is essential for users to track their condition.
- Analysis & Evaluation: Users want features that help them analyze and evaluate their own health
data, comparing current readings with historical data.
- Usability of Sensors: The ease of use, accuracy, and reliability of sensor-based functions (like
using a phone camera for pulse readings) are critical for user trust and engagement.
- System Performance & Interaction: Smooth performance, user-friendly design, and interactive
features like reminders and data sharing with medical teams are highly valued.
- Monetization Models: Users are frustrated when essential features are locked behind paywalls,
but are willing to pay for subscriptions that offer clear, substantial value.

Podcast:

About:
This study introduces ELI, an AI-based therapeutic assistant designed to enhance well-being by
providing accessible, evidence-based psychological support. Using a Design Science Research
(DSR) approach, the researchers conducted a literature review and expert workshops to establish
six core design objectives. These objectives were then implemented and evaluated in a simulated
prototype, forming the basis for future development.

Problem:
Many individuals lack timely access to professional psychological support, leading to increased
demand for digital mental health tools. However, existing AI-based solutions are often fragmented,
lack a rigorous scientific foundation, or focus narrowly on specific conditions, creating a need for a
systematically designed, reliable, and holistic therapeutic assistant.

Outcomes:
- The research established six core design objectives for an AI therapeutic assistant, focusing on
empathy, dynamic adaptability, ethical standards, technological integration, evidence-based
algorithms, and dependable support.
- A simulated prototype, ELI, was developed to instantiate these objectives, featuring a
user-friendly, speech-based interface to provide accessible, low-threshold support.
- Evaluation by 18 psychology experts found ELI to be a promising complement to traditional
therapy, particularly praising its accessibility, perceived empathy, and usefulness for milder
psychological issues.
- Key areas for improvement were identified, including the need for greater transparency in data
privacy, more robust crisis intervention capabilities, and more comprehensive therapeutic
approaches.
- While smartphones were deemed the ideal platform for accessibility, experts also saw potential for
immersive technologies like VR and AR to enhance user engagement in the future.

Podcast:

About:
This study investigates the psychological phenomenon of “AI ostracism,” where individuals feel
ignored or excluded by AI systems. The research proposes a conceptual model and a series of
online experiments to examine how providing different types of explanations for AI decisions can
reduce subsequent antisocial behavior. The core idea is that clear, actionable explanations can
restore a person's sense of control after being rejected by an algorithm.

Problem:
As AI increasingly makes critical decisions in areas like hiring and credit scoring, there's a growing
risk of negative psychological impacts on the people it rejects. While research has focused on AI
fairness and bias, little is known about the emotional and behavioral fallout of being excluded by an
impersonal system. This exclusion can diminish a person's sense of control and lead to aggressive
or antisocial responses, creating risks for individuals and organizations.

Outcomes:
- Feeling rejected by an AI system (AI ostracism) can lead to antisocial behavior, potentially more
so than rejection by a human, because it reduces a person's perceived sense of control.
- The reason for an AI's rejection matters; exclusion based on sensitive personal traits (like age or
gender) is likely to trigger a stronger negative reaction than exclusion based on qualifications.
- Providing explanations for an AI's decision can mitigate these negative effects.
- Specifically, 'counterfactual' explanations that show a user what they could have done differently
to achieve a better outcome (e.g., 'If your test score had been higher, you would have been
accepted') may be most effective at restoring control and reducing antisocial behavior.

Podcast:

About:
This study investigates why successful German mid-sized companies (Mittelstand Enterprises)
pursue incremental rather than radical digital transformation. Using a multiple-case study approach,
the research applies path dependence theory to analyze how historical success and established
business models create self-reinforcing behaviors. The analysis uncovers specific "lock-in"
mechanisms that limit companies' ability to embrace transformative digital opportunities.

Problem:
Many successful mid-sized companies are notoriously cautious about digital transformation, often
preferring small improvements over bold, innovative changes. While factors like resource
constraints are known, it's not fully understood why these firms remain on a slow trajectory even
when faced with significant digital opportunities. This research addresses why these companies
become 'trapped by their success,' leading to strategic inertia that prevents them from
fundamentally adapting their business models for the digital era.

Outcomes:
- Companies don't intentionally reject radical digital change; rather, they experience a 'functional
lock-in' where their successful, established business model reinforces a preference for incremental
improvements.
- This lock-in is driven by normative factors (ingrained company culture), cognitive biases (deeply
held assumptions about the business), and resource allocation (prioritizing existing operations over
new ventures).
- Digital initiatives are primarily adopted only if they optimize or enhance the current business
model, such as improving process efficiency or adding digital features to existing products.
- Potentially transformative digital options, like new platform-based services, are often deprioritized
or rejected because they don't align with the company's core functions and established practices.
- Without significant external pressure or a crisis, companies are unlikely to break from this
incremental path, as their current success reduces the perceived need for radical transformation.

Podcast:

About:
This study designs a visual modeling language called Workaround Modeling Notation (WAMN) to
help organizations systematically manage employee-created deviations from standard procedures.
WAMN provides a structured way to identify, visualize, and analyze the causes, conduct, and
consequences of workarounds. This enables managers to treat these informal practices not just as

Problem:
Employees frequently create 'workarounds' to overcome obstacles in their daily tasks, but these
informal solutions are often invisible to management. This lack of visibility makes it difficult for
organizations to assess their complex effects, which can range from improved efficiency to new
compliance risks and financial losses. As a result, companies struggle to make consistent, informed
decisions about how to handle these deviations.

Outcomes:
- Developed a new visual tool, WAMN, to systematically map the entire lifecycle of a workaround,
from its root cause to its positive and negative impacts.
- Enables managers to visualize complex chains of interconnected workarounds and their
cascading effects across different departments, individuals, and IT systems.
- Facilitates better decision-making by clearly showing the trade-offs (e.g., increased flexibility vs.
reduced data quality) associated with each workaround.
- Provides a framework for turning valuable employee-driven workarounds into official, improved
business processes, thereby fostering innovation from the ground up.

Podcast:

About:
This study analyzes the evolution of Information Warfare (IW) in the digital age. Using a critical
analysis of existing definitions and a hermeneutic literature review, the paper applies a digital
transformation framework to conceptualize IW as a sociotechnical process. The goal is to clarify the
ambiguous concept of IW and understand how new technologies are reshaping its methods and
societal impact.

Problem:
Information Warfare has become a powerful and pervasive threat, yet the term itself lacks a clear,
consistent definition. This conceptual ambiguity hinders the ability of organizations and
governments to develop cohesive strategies and effective policies to counter modern threats like
widespread disinformation and sophisticated cyber attacks.

Outcomes:
- Existing definitions of Information Warfare (IW) are frequently ambiguous, often confusing the
methods of attack with their outcomes, which creates a lack of conceptual clarity.
- The concept of IW has evolved significantly, shifting from a traditional military focus to a broader
societal concern that emphasizes digital technologies like AI and social media as primary enablers.
- Modern IW involves a range of attacks, including cyber espionage and psychological operations,
which can lead to significant negative societal impacts such as political polarization, reduced public
trust, and government delegitimization.
- The success of IW campaigns is heavily influenced by factors like public digital literacy and the
nature of social relationships, which can either amplify or mitigate the spread of disinformation.
- A comprehensive framework for understanding modern IW must account for the interplay between
digital technologies, social changes, and the various organizational and individual factors that
shape these conflicts.

Podcast:

About:
This study systematically reviews academic literature to categorize the essential interfaces required
for operating Data Trusts, which are organizations that manage data sharing on behalf of others. It
examines both user interfaces for human interaction and technical interfaces (APIs) for
system-to-system communication. The research provides a foundational overview of the types of
interfaces needed, identifying gaps in current knowledge to guide future development.

Problem:
Data Trusts are proposed as a key solution for trustworthy data sharing, but their practical
implementation is hindered by a lack of clarity on how users and systems should interact with them.
Without a standardized understanding of the necessary human and technical interfaces, it is difficult
to build robust, interoperable, and user-friendly Data Trusts. This gap prevents the widespread
adoption of a secure and reliable data sharing ecosystem.

Outcomes:
- Data Trusts require two main types of interfaces: Human-System interfaces for user interaction
(e.g., dashboards) and System-System interfaces for automated machine communication (e.g.,
APIs).
- Key human interfaces include web-based portals for managing consent and data access policies,
and visualization dashboards for monitoring data usage.
- Technical interfaces, particularly APIs, are crucial for enabling secure and standardized data
exchange between different organizations' IT systems.
- The research reveals a significant gap in existing literature, where interfaces are often mentioned
but lack detailed specifications or implementation guidelines for Data Trusts.
- To be effective, Data Trusts must be designed to handle various data structures (structured,
semi-structured, unstructured) and connect seamlessly to different data storage systems.

Podcast:

About:
This study analyzes the inconsistent use of hierarchical structures within taxonomies in Information
Systems (IS) research. Through a systematic literature review, the authors develop a
"meta-taxonomy"—a classification system for other taxonomies—to create a common language for
describing and understanding how hierarchical features are defined, implemented, and visualized.

Problem:
The term "hierarchical taxonomy" is applied inconsistently across IS research, leading to
conceptual ambiguity. This makes it difficult for researchers to compare studies, assess the
methodological rigor of classification systems, or select the appropriate structure for their own work,
ultimately hindering effective knowledge organization.

Outcomes:
- The research develops a preliminary meta-taxonomy to classify taxonomies that use hierarchical
features.
- This framework categorizes taxonomies based on four key dimensions: the explicitness of the
hierarchical label, the type of hierarchical relationship (e.g., parent-child), the methodological origin,
and the visualization approach (e.g., tree diagram).
- The study highlights a significant lack of methodological guidance for creating hierarchical
taxonomies, which contributes to the observed inconsistencies and provides a foundation for future
standardization.

Podcast:

About:
This study investigates how Generative AI (GenAI) is changing the creative workflows of freelance
designers. Through qualitative interviews with 10 freelance designers, the researchers performed a
thematic analysis to understand the opportunities and challenges of collaborating with text-to-image
(TTIG) AI systems.

Problem:
Generative AI is rapidly transforming creative industries, but little is understood about its specific
impact on freelance designers. These professionals are uniquely vulnerable to technological
disruption due to their direct market exposure and lack of institutional support, creating an urgent
need to understand how their work processes and job prospects are evolving.

Outcomes:
- Creativity vs. Standardization: AI can be a powerful tool for inspiration and exploring new styles,
but it also risks leading to homogenized or unoriginal creative output.
- Efficiency vs. Overprecision: While AI can drastically speed up tasks, it also tempts designers into
a cycle of endless minor tweaks and perfectionism, which can negate efficiency gains.
- AI as Sparring Partner vs. Lack of Control: Designers can interact with AI as a creative
collaborator to refine ideas, but the unpredictability and 'black box' nature of the technology can
also lead to a frustrating lack of direct control.
- Job Transition vs. Job Loss: AI is changing the designer's role towards prompt engineering and
creative direction, but it also poses a direct threat of job loss, particularly for junior freelancers
whose entry-level tasks are most susceptible to automation.

Podcast:

About:
This study analyzes the official training materials from three leading low-code/no-code (LCNC)
platforms: Appian, Mendix, and OutSystems. Using an AI-assisted method to summarize content
and extract keywords, the research assesses the feasibility of creating a standardized,
provider-independent training framework. The goal is to build a comprehensive database that can
support the development of universal training for citizen developers.

Problem:
The increasing use of low-code platforms by non-IT employees (citizen developers) requires new
training concepts to ensure application quality and security. However, current training is typically
tied to specific vendors, which creates a risk of 'vendor lock-in' and makes it difficult for companies
to build transferable skills across their workforce.

Outcomes:
- The core topics covered in the training materials from leading LCNC providers are largely
comparable and share significant overlap.
- This similarity supports the feasibility of creating a standardized, provider-independent training
framework for citizen developers.
- A standardized training approach can help organizations reduce training costs and mitigate the
risks associated with being dependent on a single vendor's platform.
- The study successfully demonstrated a partially automated, AI-driven method for analyzing and
comparing training content across different platforms.

Podcast:

About:
This study investigates how mobile applications can use psychological principles to effectively
encourage sustainable consumer behavior. Researchers first analyzed 31 sustainability apps to
identify the most commonly used psychological factors based on the SHIFT framework.
Subsequently, an online experiment was conducted to empirically test the effectiveness of the two
most prevalent factors in promoting eco-friendly actions and attitudes.

Problem:
While individual lifestyles are a major contributor to climate change, many people fail to translate
their sustainable intentions into action, a problem known as the 'attitude-behavior gap'.
Sustainability mobile apps are designed to address this, but they often suffer from high user dropout
rates, indicating they may not be effectively motivating long-term behavioral change.

Outcomes:
- The most frequently used psychological strategies in sustainability apps are making sustainable
actions easy ('Habit formation') and providing clear information and knowledge ('Feelings and
cognition').
- Experimental results confirmed that both strategies are highly effective in positively influencing
users' eco-conscious behavior, sense of obligation, and attitude towards sustainability.
- Combining both strategies ('making it easy' plus providing 'information') did not produce a
significantly stronger effect than using just one of the strategies alone.
- These psychological triggers in mobile apps can successfully motivate users to adopt more
sustainable habits, helping to close the gap between environmental intention and action.

Podcast:

About:
This study investigates the use of Large Language Models (LLMs) to automatically extract and
classify "explanatory rationales"—the underlying reasons for specific sequences of activities in a
business process—from textual documents. The researchers compared various LLMs and four
different prompting techniques to evaluate their effectiveness on a real-world use case involving a
university thesis process.

Problem:
Understanding why business process steps occur in a certain order (e.g., due to law, internal policy,
or best practice) is crucial for process improvement and compliance. However, this information is
typically extracted manually from documents and interviews, which is a slow, expensive, and
labor-intensive task. This research addresses the need for an automated method to identify and
categorize these activity relationships.

Outcomes:
- Large Language Models (LLMs) can successfully automate the extraction and classification of the
reasons behind business process rules from text.
- Providing a few examples in the prompt ('Few-Shot learning') dramatically improves the accuracy
of the LLMs compared to basic instructions.
- Smaller, more cost-effective models like GPT-4o mini can achieve results comparable to or even
better than larger, more expensive models when using appropriate prompting techniques.
- This automation makes advanced process analysis more accessible to organizations with limited
resources, helping to improve compliance checking and business process redesign efforts.

Podcast:

About:
This study investigates how organizations can effectively build the necessary skills and capabilities
for digital transformation. Through a qualitative case study of a large media and technology
company, the research identifies a clear, sequential process for developing digital competence,
offering practical insights for business leaders.

Problem:
Many companies struggle with digital transformation because existing guidance on building the right
employee skills is often too abstract and lacks concrete examples. This research addresses the gap
by providing a practical, real-world roadmap for how a company successfully developed its digital
competencies over time.

Outcomes:
- Building digital competence is a three-stage journey: 1) Expand core IT skills, 2) Create an agile
and supportive organizational structure (Meta Competence), and 3) Foster the ability to innovate
new products and business models (Transformation Competence).
- Different types of competence require different development tools. Core IT skills can be built with
traditional training, but creating an agile culture requires organizational changes like new
management roles, faster decision-making, and greater employee involvement.
- Fostering true innovation skills requires moving beyond rigid classroom training. Companies
should use methods that encourage experimentation, self-directed learning, and practical
application, such as product development events and hackathons.
- A supportive, flexible organizational environment is crucial. It acts as a foundation that enhances
the development of both technical skills and broader innovative capabilities.

Podcast:

About:
This study analyzes when predatory pricing is a rational strategy for competitive digital platforms.
Using a multi-stage game theory model that incorporates uncertainty and deep reinforcement
learning, the researchers simulate platform competition to identify the conditions that lead to market
monopolization.

Problem:
In the digital economy, markets often consolidate into monopolies even when competing platforms
appear evenly matched. Traditional economic models struggle to explain this phenomenon because
they often assume competitors have complete information about each other's costs, which is
unrealistic and overlooks the role of uncertainty in strategic pricing.

Outcomes:
- Uncertainty about a competitor's costs is a primary driver of monopolization. Monopolies emerged
in roughly 60% of simulated cases, even when platforms were otherwise identical.
- In contrast, when competitors have complete information, monopolies only tend to form when one
platform has a clear cost advantage.
- Asymmetries, such as one platform having slightly lower costs, significantly increase the likelihood
that the market will become a monopoly.
- Risk aversion among managers has the opposite effect, making aggressive predatory pricing less
likely and reducing the chance of monopoly formation.

Podcast:

About:
This study investigates gender bias in Large Language Models (LLMs) like ChatGPT within the
context of digital innovation and entrepreneurship. Researchers used two tasks—associating
professions with gendered terms and simulating venture capital (VC) funding decisions—to identify
if and how AI outputs reflect and reinforce societal gender stereotypes.

Problem:
As businesses rapidly adopt AI tools for tasks ranging from brainstorming to decision-making,
there's a significant risk that these technologies could perpetuate harmful gender biases. This can
create disadvantages for women in business and tech, potentially hindering innovation and
reinforcing existing inequalities in professional fields.

Outcomes:
- AI reflects societal biases: ChatGPT-4o was more likely to associate professions in digital
innovation and tech with men, mirroring real-world gender disparities.
- AI can simulate exclusionary behavior: In a simulated VC funding scenario, the AI model showed
'in-group bias,' where male VCs preferred male entrepreneurs and female VCs preferred female
entrepreneurs, reinforcing stereotypes in professional decision-making.
- Bias mitigation is complex: Some AI models attempt to counteract bias but can overcorrect,
leading to unrealistic outcomes that mask the underlying problem rather than solving it.
- Implicit bias is a key risk: The study found that bias was present even when gender was only
implied through names, highlighting how AI can perpetuate stereotypes in subtle ways that are
difficult for users to detect.
- Businesses must be cautious: Companies using AI for ideation, hiring, or investment analysis
need to be aware that these tools can introduce or amplify bias, potentially leading to unfair
outcomes and overlooked talent.

Podcast:

About:
This study develops an artificial intelligence system to optimize postal and newspaper delivery
routes in Würzburg, Germany. The approach uses a Graph Neural Network (GNN), a type of deep
learning, to automatically divide a city into more efficient and balanced delivery districts. The
system's performance was evaluated using real-world census and geographic data to demonstrate
its potential advantages over existing methods.

Problem:
Logistics companies often struggle to create fair and efficient delivery zones, leading to imbalanced
workloads, higher operational costs, and lower employee satisfaction. Current AI solutions for this
'districting and routing' problem are often too rigid, failing to account for real-world complexities like
variable customer density or the need for equitable route lengths among delivery staff.

Outcomes:
- The proposed AI prototype consistently created more cost-efficient delivery districts than two
standard benchmark models, reducing overall travel time.
- Despite its efficiency, the model produced highly imbalanced workloads; in one test, the longest
delivery route was more than four times the length of the shortest one.
- The research confirms that deep learning holds significant potential for solving complex districting
problems, but highlights the critical need to incorporate workload balancing as a primary objective
for practical, real-world applications.

Podcast:

About:
This study investigates how the stock market reacts to announcements of digital platform
acquisitions. Using an event study methodology on a global sample of 157 firms, the research
analyzes how factors like the strategic motivation for the acquisition (innovation vs. efficiency) and
the maturity of the target platform influence investor perception and the acquiring firm's value.

Problem:
Acquiring digital platforms is a popular but risky growth strategy, and little is known about its actual
effectiveness. Companies and investors face uncertainty regarding the value of these deals due to
high costs, complex integration challenges, and difficulties in realizing synergistic benefits, creating
a need for clear guidance on successful acquisition strategies.

Outcomes:
- Announcing the acquisition of a digital platform generally leads to a negative stock market reaction
for the acquiring firm, signaling investor concerns.
- The negative market impact is significantly reduced when the acquisition is 'exploration-oriented'
(focused on innovation and long-term growth) rather than 'exploitation-oriented' (focused on
immediate efficiency gains).
- Acquiring a mature platform with an established user base and proven network effects results in a
more favorable, and sometimes positive, market reaction compared to acquiring a nascent (new)
platform.

Podcast:

About:
This study explores using Large Language Models (LLMs) to automatically translate diverse
patient-generated health data (PGHD) from sources like wearables into the standardized FHIR
format. The researchers developed and tested a data mediation pipeline that combines an LLM with
a validation and correction mechanism to ensure the output data is usable in clinical systems. The
study evaluated two different prompting strategies across various data formats, such as JSON,
CSV, and free text, to assess the pipeline's effectiveness.

Problem:
Integrating health data from various patient wearables and devices into hospital information
systems is a significant challenge due to a lack of data standardization. This poor interoperability
results in fragmented health information, which can lead to misinformed medical decisions and
hinder effective patient care. Current methods for standardizing this data are often manual,
labor-intensive, and struggle to keep up with the fast-paced market of wearable technology.

Outcomes:
- Large Language Models (LLMs) can effectively automate the conversion of patient-generated
health data from multiple formats into the standardized FHIR format, greatly improving healthcare
data interoperability.
- A data pipeline that includes a validation and self-correction loop significantly increases the
reliability and accuracy of the LLM's output, ensuring it meets FHIR standards.
- The study found that providing the LLM with a few concrete examples ('few-shot' prompting) was
significantly more effective than giving it abstract rules ('reasoning' prompting).
- While successful in reformatting data, the LLM struggled with semantic accuracy for complex
structured data (like JSON or CSV) that required mathematical aggregation, such as summing up
sleep stage durations.
- The system performed best with unstructured free-text data, indicating LLMs are highly adept at
processing natural language but face challenges with complex, nested data structures that require
calculations.

Podcast:

About:
This study performs a meta-analysis, synthesizing data from 46 previous studies involving over
21,000 participants, to understand how people perceive the justice and fairness of algorithmic
decision-making (ADM). The research systematically compares these perceptions against those of
traditional human decision-making to identify overall trends and influential factors.

Problem:
As businesses increasingly adopt algorithms for critical decisions in areas like hiring and
management, there is conflicting evidence on how these systems are perceived by individuals. This
lack of clarity makes it difficult for organizations to predict whether implementing ADM will be
viewed as fair or if it will lead to negative reactions and consequences.

Outcomes:
- On average, people perceive decisions made by algorithms as less fair and less just compared to
decisions made by humans.
- These negative perceptions of algorithmic decision-making (ADM) lead to unfavorable outcomes
for organizations, such as reduced attractiveness to potential employees.
- The context where ADM is applied is critical; while it is viewed negatively in recruiting, it can be
perceived positively in other areas like education and leadership.

Podcast:

About:
This study investigated employee acceptance of metaverse technologies within the traditionally
conservative paper and packaging industry. Using a 'living lab' experiment at a leading company,
researchers used surveys and observation to identify the key factors influencing whether
employees would be willing to adopt immersive virtual tools for their work. The analysis was based
on the well-established Technology Acceptance Model 3 (TAM3).

Problem:
As companies invest heavily in the metaverse to improve remote collaboration and efficiency, there
is a significant knowledge gap regarding employee acceptance of these new technologies. It is
unclear what drives or prevents adoption in real-world business environments, particularly in
industrial sectors that are typically slower to digitize. This research addresses this gap by exploring
the practical barriers and benefits from the employee's perspective.

Outcomes:
- Employee adoption of the metaverse hinges on two main factors: its perceived usefulness (PU) for
their job and its perceived ease of use (PEU).
- The technology's relevance to an employee's specific daily tasks is the single most important
factor driving its perceived usefulness.
- An employee's confidence with computers (computer self-efficacy) is the strongest predictor of
whether they will find the metaverse easy to use.
- Overall acceptance was moderate (3.61 out of 5), with employees seeing potential for
collaboration but citing concerns over immature technology and clunky hardware like heavy VR
headsets.
- Lack of prior experience was a significant barrier, as many participants struggled to imagine
practical applications for the metaverse in their day-to-day work.

Podcast:

About:
This study investigates how university students who also work part-time use Generative AI (GenAI)
in both their academic and professional lives. Using in-depth interviews with eleven students from a
distance learning university, the research employs a grounded theory approach to build a model
explaining the factors, strategies, and consequences of their GenAI usage.

Problem:
While the use of GenAI like ChatGPT is widespread, little research has focused on the unique
challenges and behaviors of part-time students who must navigate its application across both
educational and business settings. This study addresses the gap by exploring how this specific
group balances the productivity benefits of GenAI against concerns like data privacy, accuracy, and
academic integrity in their dual roles.

Outcomes:
- Productivity Boost: Students use GenAI to increase efficiency and save time on tasks for both
school and work, such as summarizing content, adapting language, and generating ideas.
- Cautious Application: While leveraging GenAI for professional tasks, students are highly aware of
data security risks and avoid inputting sensitive company information into public tools.
- Key Concerns Remain: The primary drawbacks identified are the risk of inaccurate or
'hallucinated' outputs, the need for clear ethical guidelines from universities, and privacy
considerations.
- Skill for the Future: Participants view proficiency in GenAI as a critical future job skill and
expressed a strong desire for more formal training from their universities.
- Replacing Traditional Tools: GenAI is frequently used as a faster alternative to traditional search
engines like Google for information gathering.

Podcast:

About:
This study explores how Algorithmic Management (AM) is implemented in hospitals to handle
coordination and control tasks traditionally performed by human managers. Using a qualitative
approach, the research conducted nine semi-structured interviews with doctors, software providers,
and a domain expert to identify how AM is applied across the hospital value chain, from patient
admission to administration.

Problem:
While Algorithmic Management is well-studied in low-skill, platform-based work, its application in
high-skill, traditional industries like healthcare remains largely unexplored. Hospitals face significant
challenges, including workforce shortages and complex coordination demands, creating an urgent
need to understand how algorithmic systems can improve operational efficiency and resource
allocation in this high-stakes environment.

Outcomes:
- The study identified five key use cases for algorithmic management in hospitals: patient intake
management, bed management, doctor-to-patient assignment, workforce management, and
performance monitoring.
- AM systems significantly improve hospital efficiency by automating administrative tasks, reducing
staff workload, optimizing bed occupancy, and ensuring fairer task distribution among doctors.
- Unlike in the gig economy, AM in hospitals complements human managers by embedding
algorithmic logic into existing structures rather than replacing them entirely.
- While beneficial for efficiency, AM can also create pressure on staff through rigid schedules and
time-based performance targets, and performance monitoring often lacks transparency for
clinicians.
- The technology enhances patient care by algorithmically prioritizing urgent cases, matching
patients with the most suitable doctors, and improving continuity of care for follow-up visits.

Podcast:

About:
This research introduces the concept of a Patient Pathway Template Repository (PPTR) to
standardize and improve cancer care processes within European healthcare networks. The study
lays the groundwork for developing this repository by using a Design Science Research approach,
which involves analyzing stakeholder expectations and reviewing existing literature. The ultimate
goal is to create a central platform for the development, management, and implementation of
patient pathway templates.

Problem:
Developing and implementing patient pathways—structured plans for patient care—is a complex,
costly, and non-standardized task, especially in multifaceted areas like cancer treatment. This lack
of a unified approach leads to inconsistencies in care quality and inefficient use of resources across
different hospitals and regions. This study addresses the absence of a shared, collaborative
platform for creating, adapting, and exchanging evidence-based patient pathway templates among
healthcare providers.

Outcomes:
- Stakeholders anticipate that a central repository will significantly improve the standardization and
quality of cancer care, leading to better patient outcomes.
- The main benefits identified are enhanced knowledge sharing, the ability to compare performance
against best practices, and increased efficiency by avoiding the duplication of effort.
- Key concerns raised by potential users include the practical usability of templates in daily clinical
routines, their adaptability to local needs, and the challenges of long-term maintenance and
updates.
- The research concludes that a successful repository must be user-friendly, allow for both
standardization and flexibility, and be supported by a community-driven approach and a clear
governance model.

Podcast:

About:
This study explores how to improve digital inclusion for senior citizens by combining user interface
design guidelines with process guidance. Using a Design Science Research approach, the
researchers developed and iteratively refined a prototype of a travel booking website based on
heuristic evaluations with 13 senior participants.

Problem:
As essential services and business processes increasingly move online, senior citizens often face
significant barriers to digital participation, leading to technological and social disadvantages. This
creates a digital divide that excludes a growing segment of the population, representing a missed
opportunity for businesses to expand their customer base.

Outcomes:
- A structured process guidance system, which visually maps out all steps of an online task,
significantly improves performance and reduces completion time for senior users.
- Key design principles for seniors include reducing complexity, using large fonts with high contrast,
maintaining design consistency, and providing clear, immediate feedback.
- Integrating help instructions directly into the interface is more effective than separate help
sections, as it lowers the cognitive load on users.
- Businesses that adapt their online interfaces to be senior-friendly can better serve a growing
demographic, enhancing both social inclusion and commercial reach.

Podcast:

About:
This research identifies the critical factors for successfully scaling smart city (SC) pilot projects
beyond the experimental stage. Through a thematic analysis of 129 academic articles, the study
develops a holistic framework that categorizes key success factors into four dimensions: financial,
organizational, social, and technical. The goal is to provide a comprehensive guide for cities to
overcome common barriers and realize the full potential of their SC initiatives.

Problem:
Many promising smart city projects fail to progress after a successful trial, a problem known as the
'piloting trap'. This is often caused by a lack of economic sustainability, viable operational models,
and market readiness. Existing research on this issue is frequently limited to specific use cases or
regions, highlighting a need for a more universal framework to guide successful scaling across
different urban contexts.

Outcomes:
- The successful upscaling of smart city projects requires a holistic approach that balances four
interdependent dimensions: Financial, Organizational, Social, and Technical.
- Key financial drivers include securing institutional funding, creating clear monetization plans, and
conducting thorough risk assessments and economic value calculations.
- Organizational success is most dependent on strong inter-organizational collaboration and
partnerships, supported by effective governance, leadership, and knowledge transfer.
- Social acceptance hinges on high levels of citizen engagement and participation, as well as
establishing trust and ensuring the reliability of the new systems.
- From a technical perspective, interoperability between different systems is the single most critical
factor for seamless data exchange and cost-effective scaling.

Podcast:

About:
This study investigates how to design a Digital Service Innovation Hub (DSIH) to help
small-to-medium enterprises (SMEs) and public organizations collaborate on innovation. Using
insights from 17 expert interviews and focus groups, the research identifies common challenges
and proposes a detailed set of requirements for building an effective hub that actively coordinates
innovation activities.

Problem:
Small and medium-sized enterprises (SMEs) and public sector organizations often struggle to
develop new digital services because they lack the necessary resources, expertise, and
connections. Existing innovation hubs typically focus on narrow technological solutions and fail to
provide the comprehensive, coordinated support needed to navigate complex, multi-partner
innovation projects.

Outcomes:
- The study identified four key barriers to innovation for SMEs and public organizations: external
factors (market and technology shifts), internal challenges (resistant culture, resource shortages),
gaps in knowledge and digital skills, and difficulties in finding and managing partners.
- To overcome these barriers, a Digital Service Innovation Hub (DSIH) should be designed around
three core functions: orchestrating the collaboration of different actors, facilitating knowledge
transfer, and ensuring the platform is practical and sustainable.
- As an 'orchestrator,' the hub must provide advanced tools for matchmaking partners (e.g.,
AI-based recommendations and search filters), mediate collaborations, and clearly communicate its
value to attract users.
- The hub must act as a learning catalyst by providing access to curated content, best practices,
and success stories, as well as structured formats like forums and mentoring for peer-to-peer
knowledge sharing.
- For successful adoption, the DSIH platform must be user-friendly, offer modular solutions, provide
reliable technical support, and operate with a transparent business model that defines engagement,
pricing, and expectations.

Podcast:

About:
This study investigates how Transactive Memory Systems (TMS), a theory explaining how human
teams manage shared knowledge, manifests in human-Generative AI (GenAI) collaboration.
Through qualitative interviews with 14 knowledge workers, the research analyzes the unique
dynamics of specialization, credibility, and coordination that emerge when humans partner with AI.
The findings reveal significant deviations from traditional human-human collaboration patterns.

Problem:
As GenAI is increasingly integrated into knowledge work as a collaborative partner, existing
frameworks for understanding teamwork are becoming outdated because they were designed for
human-only interactions. This creates a gap in understanding how to effectively structure human-AI
partnerships, often leading to sub-optimal workflows, imbalanced responsibility, and misplaced
trust.

Outcomes:
- Humans exhibit ambivalent trust in GenAI, recognizing its expertise but simultaneously
questioning its credibility, which necessitates constant manual verification of its output.
- Collaboration is asymmetrical and hierarchical, with humans consistently initiating tasks and
directing a passive GenAI that lacks initiative, in contrast to the proactive and reciprocal nature of
human teams.
- Expertise recognition is one-sided; while humans learn GenAI's capabilities, the AI fails to develop
a persistent memory of the human's skills or context, resetting its understanding with each new
interaction.
- The study proposes adapting teamwork theories to account for GenAI's temporary memory and
lack of a social dimension to design better collaborative systems.

Podcast:

About:
This study investigates citizens' perceptions of social risks associated with smart city development.
Through a quantitative survey of 310 participants in Germany and Italy, the research identifies 15
key social risks and analyzes how citizens rate their probability of occurrence and potential severity.

Problem:
The rapid digital transformation of cities often focuses on technological benefits while overlooking
potential social dangers like privacy violations, increased surveillance, and cybersecurity threats.
This gap can undermine public trust and well-being, yet citizens' perspectives on these risks are
frequently ignored during the planning and implementation of smart city initiatives.

Outcomes:
- Citizens have a dual perception: they maintain a generally positive attitude towards smart cities
but also rate the associated social risks as highly probable and severe.
- The risk of personal 'profiling' was perceived as the most probable, while 'cybersecurity threats'
were seen as having the most severe impact.
- Risk perception is influenced by demographics, with older participants and Italian citizens rating
risks higher than their younger and German counterparts.
- Despite high concern for risks, citizens are willing to engage in and cooperate with smart city
projects, indicating a desire for more participatory and ethical development.

Podcast:

About:
This study analyzes the current state of retail service robot (RSR) adoption in physical,
brick-and-mortar (B&M;) stores. Using a dual-method approach, the research combines a
systematic review of academic literature with a multi-case study of leading European retailers. The
goal is to provide a holistic overview of how robots are being used for specific operational tasks in
theory versus practice.

Problem:
Brick-and-mortar retailers face growing pressure from acute staff shortages, rising operational
costs, and intense competition from e-commerce. While retail service robots offer a potential
solution to improve efficiency and customer experience, there is a lack of clear understanding of
how they are currently being implemented and which tasks they perform, creating a gap between
their potential and actual use.

Outcomes:
- Retail service robots are most commonly adopted for customer service tasks, information
exchange, and transporting goods within the store.
- The full potential of RSRs is largely untapped, as they are typically used for simple, single-purpose
tasks rather than complex or multi-functional roles.
- A significant gap exists between academic research, which heavily focuses on customer-facing
robots, and real-world deployment, where retailers are more cautious about using robots in
customer service.
- Current RSR development prioritizes customer interactions, with much less consideration given to
how robots can support employees and integrate into their daily workflows.
- Overall, RSR adoption in retail is still in its early stages, hindered by organizational barriers,
technological challenges, and the need for greater understanding of both customer and employee
acceptance.

Podcast:

About:
This study analyzes existing methods for identifying and labeling AI-generated deepfakes in the
context of new European Union regulations. Through a comprehensive literature review, the
authors assess the limitations of individual detection techniques. They then propose a novel
multi-level strategy that combines technical detection, trusted human review, and risk assessment
to help online platforms moderate content effectively and at scale.

Problem:
The proliferation of deepfakes poses a significant threat to society by enabling the spread of
misinformation and manipulating public opinion, especially in politics. While the EU has enacted
new laws like the AI Act and Digital Services Act to mandate transparency, online platforms lack a
practical and scalable framework to comply. Individual detection methods are often insufficient to
handle the volume and sophistication of modern deepfakes, creating a critical enforcement gap.

Outcomes:
- Existing single methods for detecting deepfakes, such as watermarking or artifact analysis, are
inadequate on their own to meet the demands of EU regulations and are vulnerable to
manipulation.
- The research proposes a practical multi-level strategy that combines multiple detection
approaches to create a more robust content moderation system for online platforms.
- The first level of the strategy quickly categorizes content by checking for embedded digital
markers that either identify it as a deepfake or verify it as authentic.
- For unmarked content, a second level uses a scoring system that combines automated technical
detection, trusted human verification, and an assessment of the content's potential for harm.
- This risk-based scoring allows platforms to apply clear labels such as 'Deepfake', 'Warning', or
'Verified', making the moderation process more transparent, scalable, and adaptable.

Podcast:

About:
This study investigates how employees perceive the fairness of decisions made with AI
involvement, specifically comparing two approaches: full delegation to an AI versus a collaborative
human-AI 'ensemble'. Through an online experiment involving 79 participants, the research
measured how these different decision-making processes, in the context of a performance
evaluation, affect perceptions of fairness and trust in management.

Problem:
As companies increasingly use AI for important decisions to improve efficiency, they risk negative
reactions from employees. Decisions involving AI are often perceived as less fair than those made
by humans, which can erode trust in leadership and harm morale. This study addresses how
different methods of integrating AI can either worsen or mitigate these negative perceptions.

Outcomes:
- Fully delegating a decision, such as a performance review, to an AI system is perceived as
significantly less fair than a decision made by a human manager.
- This reduced perception of fairness leads directly to lower employee trust in the manager who
delegated the task.
- Using a collaborative 'ensemble' approach, where a manager and an AI are equally involved in the
decision, does not cause this negative effect; it is perceived as just as fair as a human-only
process.
- The outcome of the decision (e.g., a high vs. low bonus) does not change the negative perception
of unfairness when a task is fully delegated to an AI.
- To maintain trust, businesses should favor collaborative human-AI models over complete
delegation to autonomous systems.

Podcast:

About:
This study investigates how blockchain verification and the type of issuing institution (university vs.
learning academy) impact employer perceptions of a job applicant's trustworthiness, expertise, and
salary expectations. Using an experimental design, 200 participants evaluated a candidate based
on one of four micro-credential variations to measure differences in hiring decisions.

Problem:
Traditional academic credentials are slow and costly to verify, while new micro-credentials from
various providers suffer from a lack of trust and recognition among employers. This study
addresses the uncertainty of whether blockchain technology can bridge this trust gap and make
micro-credentials a viable and valued tool in the hiring process.

Outcomes:
- Blockchain verification did not significantly increase employers' perception of an applicant's
trustworthiness or expertise.
- Employers showed no significant preference for credentials issued by a university over those from
a non-university learning academy, suggesting a shift toward skills-based evaluation over
institutional prestige.
- Applicants with blockchain-verified credentials paradoxically received lower minimum salary
offers, possibly because easily verifiable credentials reduce perceived hiring risk, leading to more
conservative initial offers.
- The findings indicate that the source of a credential is becoming less important to employers than
the demonstrated skills it represents.

Podcast:

About:
This study introduces "ConSight," a prototype system designed to help knowledge workers in the
construction industry manage and retrieve information from large, unstructured document
collections. Using a Design Science Research methodology, the system leverages Large Language
Models (LLMs) to integrate information retrieval with multi-modal data support, enhancing
information access and providing contextual understanding.

Problem:
Knowledge workers, particularly in document-heavy sectors like construction, are often
overwhelmed by the sheer volume of unstructured data they must process daily. Current
LLM-based systems struggle to provide accurate, relevant, and context-specific information
efficiently, often failing to handle diverse data formats beyond simple text or ensure the
trustworthiness of their outputs.

Outcomes:
- Developed 'ConSight,' a prototype system that uses Large Language Models (LLMs) to help
professionals extract and structure information from large, unstructured document collections.
- Identified eight key design requirements for an effective knowledge support tool, including
multi-perspective document organization, user-specific information displays, timeline views, and the
detection of data inconsistencies.
- Designed a modular architecture with features for adaptive intelligence, document management,
and validation, ensuring that all AI-generated insights are traceable back to their original source.
- Created a user-facing dashboard that organizes complex information by topic or timeline,
enhancing accessibility and building user trust through transparency.
- The research provides a practical framework for building AI tools that can turn massive document
repositories into actionable, contextual insights.

Podcast:

About:
This study investigates how first-semester university students perceive feedback on their
programming projects generated by Large Language Models (LLMs) compared to feedback from
their peers. The research involved 167 students in an introductory programming and data science
course who evaluated the LLM feedback on criteria such as helpfulness, clarity, and motivational
quality.

Problem:
Providing personalized, high-quality feedback in large educational courses is challenging due to
limited instructor time and resources. While peer feedback is a common alternative, its
effectiveness can be undermined by the limited expertise of student reviewers and a reluctance
from others to accept peer judgments, creating a need for scalable and effective feedback
solutions.

Outcomes:
- A majority of students (56%) preferred the LLM-generated feedback over the feedback provided
by their peers.
- Students struggled to differentiate between AI and human feedback, with 52% unable to correctly
identify which comments were written by the LLM.
- The LLM feedback was perceived positively across most quality dimensions, including being
constructive, helpful, understandable, and satisfying.
- The main criticisms were that the feedback lacked a motivating element and students wished it
were more personalized and offered more detailed suggestions for improvement.

Podcast:

About:
This study addresses the challenges small and medium-sized enterprises (SMEs) face in using
maturity models (MMs) to assess their business capabilities. Researchers analyzed 28 academic
papers to develop ten specific design principles for creating more usable and effective MMs tailored
to SMEs. The practical usefulness of these principles was then confirmed through a survey of 18
industry and academic experts.

Problem:
Standard maturity models, which help organizations assess and improve their processes, are often
too complex, resource-intensive, and generic for small and medium-sized enterprises (SMEs). This
misalignment prevents SMEs, which have limited resources and unique needs, from effectively
using these valuable tools to guide their growth and digital transformation.

Outcomes:
- The study produced ten actionable design principles for developing SME-focused maturity
models.
- These principles emphasize creating models that are adaptable to specific company sizes or
sectors and provide a holistic but simple view of capabilities.
- A key recommendation is to include intuitive self-assessment tools, reducing the need for costly
external consultants.
- The principles advocate for a balance between theoretical soundness and practical, real-world
applicability.
- An expert survey confirmed the high practical value of the proposed principles, indicating they can
successfully guide the creation of more effective tools for SMEs.

Podcast:

About:
This study analyzed customer interactions with a humanoid robot named Pepper, which was
deployed in a German bank branch for two weeks. The research aimed to understand how the
robot affected customers' perception of service quality by observing its performance in two roles:
assisting with online banking troubleshooting and providing entertainment.

Problem:
The banking industry, particularly in Germany, is facing a significant labor shortage, which
threatens the quality of in-person customer service. Banks are exploring automation like service
robots to handle routine inquiries, improve operational efficiency, and free up human employees for
more complex financial advisory tasks.

Outcomes:
- The robot failed at its primary task of providing online banking support, succeeding in only 7% of
attempts due to technical limitations and customer impatience.
- The robot was highly successful as an entertainment and information tool, completing 76% of
these interactions and eliciting positive emotional responses from customers.
- The robot's presence significantly enhanced the bank's innovative image, serving as a
public-facing symbol of technological adoption.
- Customers strongly preferred using the robot's touchscreen over verbal commands, suggesting
that familiar interfaces are more effective for customer-robot interaction.
- The robot had a positive social effect, attracting attention and sparking conversations among
customers in the bank branch.

Podcast:

About:
This study investigates the trade-offs consumers make when choosing smart home devices. Using
a choice-based conjoint analysis, the research evaluates the relative importance of eight different
attributes related to performance, privacy, and market factors like price and provider.

Problem:
As smart home technology becomes more common, it's unclear how consumers weigh competing
factors like device performance, data privacy, and price. This study addresses the gap in
understanding these decision-making trade-offs, which is crucial for designing products that
consumers will trust and adopt.

Outcomes:
- Reliability is the most important factor for consumers when choosing a smart home device,
followed closely by the provider's reputation and location.
- Surprisingly, price and specific privacy features (like data storage location or user controls) play a
much smaller role in the decision-making process than reliability and provider.
- Consumers strongly prefer devices from domestic (German) providers and are wary of storing
their data in the cloud, preferring local storage on the device itself.
- While consumers are concerned about data collection, they are willing to share data for product
improvement but not for revenue generation or marketing purposes.

Podcast:

About:
This study conducts a systematic literature review to examine how Large Language Models (LLMs)
can advance Intelligent Automation (IA). The research analyzes existing academic papers to
categorize the functional roles of LLMs in automation and identifies key capabilities, current
applications, and significant research gaps.

Problem:
Traditional Robotic Process Automation (RPA) is limited in its ability to handle unstructured data,
adapt to changing workflows, and perform complex reasoning. While LLMs show promise in
overcoming these limitations, there has been no systematic investigation into how they can be
effectively integrated to enhance Intelligent Automation (IA).

Outcomes:
- LLMs are primarily used in Intelligent Automation in three distinct roles: as specialized 'Tools' to
process complex inputs, as a means 'to Build IA' by generating automation workflows from natural
language, and as 'Agents' to dynamically navigate software and make decisions.
- The integration of LLMs makes business process automation significantly more flexible and
robust, allowing systems to handle complex data and adapt to changes in software interfaces,
which is a major weakness of traditional RPA.
- A crucial research gap was identified in the current approaches: the lack of systems that can learn
from feedback at runtime. Most implementations deploy LLMs as static models, missing a core
component of true intelligence.

Podcast:

About:
This study proposes a framework to understand how Virtual Reality (VR) can be used effectively in
online retail. It investigates how the key VR features of immersion and interactivity influence a
customer's psychological state and their ultimate intention to purchase a product. The research
outlines an experimental methodology to test how different levels of these features impact
consumer behavior.

Problem:
Retailers see great potential in using VR to create unique and engaging shopping experiences, but
its actual deployment remains limited. There is a lack of clear understanding of how to best design
VR product presentations to reliably influence consumer perceptions and drive sales. This study
addresses the gap by examining the specific psychological mechanisms that make VR an effective
sales tool.

Outcomes:
As a 'Research in Progress' paper, this study presents a theoretical model and hypotheses rather
than final results. The key proposals are:
- More immersive and interactive VR presentations will increase customer engagement, curiosity
(enticement), and confidence in evaluating the product.
- Providing complete interactivity is expected to maximize user engagement, while strategically
restricting interaction might be better for sparking curiosity.
- Positive changes in engagement, curiosity, and evaluation confidence are all predicted to directly
increase a customer's intention to buy the product.

Podcast:

About:
This study introduces an efficient method to detect incorrectly labeled data in industrial quality
control systems, specifically for automated surface inspection in steel production. The researchers
adapted a technique called Area Under the Margin (AUM) ranking to work with gradient-boosted
decision tree models, a common tool for analyzing tabular data. The approach was validated on
several datasets, including real-world industrial data, to demonstrate its ability to improve machine
learning data quality.

Problem:
In manufacturing, machine learning systems are used to automatically detect surface defects, but
their accuracy depends heavily on correctly labeled training data. Errors in this data, caused by
human mistakes or ambiguous defect types, can significantly degrade the system's performance
and reliability. This study addresses the lack of efficient, practical methods for finding these label
errors specifically within the tabular data formats commonly used in industrial quality control.

Outcomes:
- The proposed Area Under the Margin (AUM) method effectively identifies incorrectly labeled data
in real-world industrial datasets, such as those from steel manufacturing.
- The technique is computationally efficient, requiring only a single training run of the machine
learning model, making it faster and more practical than many existing alternatives.
- Its performance in detecting label errors is competitive with more complex and computationally
intensive state-of-the-art methods.
- The AUM score was successfully integrated into an industrial quality control workflow, significantly
reducing the manual effort and time required for experts to find and correct data errors.
- Cleaning training data by removing the samples identified by the lowest AUM scores was shown
to improve the overall accuracy of the defect classification models.

Podcast:

About:
This paper provides a systematic literature review of Self-Sovereign Identity (SSI), analyzing 78
academic articles to identify key challenges and requirements from a sociotechnical perspective.
The study maps the current research landscape, highlighting the dominant focus on technical
aspects and the neglect of social factors. The goal is to understand what is needed for the
successful implementation and widespread adoption of SSI.

Problem:
As individuals use more internet services, they lose control over their personal data, which is
collected and managed by large tech corporations. Self-Sovereign Identity (SSI) is a proposed
solution to give users back control, but academic research has largely overlooked the social
challenges to its adoption, such as user acceptance and usability, focusing instead almost
exclusively on technical issues like security.

Outcomes:
- Academic research on SSI is overwhelmingly focused on technical aspects like security and
privacy, while crucial social factors such as user acceptance, trust, and usability are significantly
underrepresented.
- Many SSI implementations rely on blockchain technology, which paradoxically presents
challenges to privacy and scalability, two core requirements for a successful identity system.
- Most research treats SSI in a general sense, with a lack of studies focused on specific application
domains like healthcare or e-government, where unique requirements and challenges exist.
- Social factors are critical for the success of any SSI solution, as adoption depends heavily on user
experience, trust in the system, and perceived control.
- The study identifies key application areas for SSI in e-government, finance, healthcare,
e-commerce, and the Internet of Things (IoT).

Podcast:

About:
This study investigates how future professionals develop proficiency with artificial intelligence (AI).
Through a survey of 352 business school students, the researchers examined whether hands-on AI
experience (both using and designing systems) improves AI literacy by first building a solid
foundation of theoretical AI knowledge.

Problem:
As AI becomes integral to modern jobs, professionals must be 'AI literate' to use these tools
effectively and responsibly. However, organizations and educators lack a clear model for how this
skill actually develops, often assuming that simple usage is enough. This research addresses that
gap by testing if structured knowledge is a necessary stepping stone between experience and true
proficiency.

Outcomes:
- Hands-on experience is most effective when it builds a person's structured knowledge about AI's
principles, processes, and ethical considerations.
- This foundational AI knowledge is the critical bridge that translates hands-on experience into
practical AI proficiency (literacy).
- Experience involving designing or configuring AI systems, even at a basic level, has a stronger
impact on developing AI literacy than just using AI tools.
- For businesses and educators, effective AI training must combine practical application with formal
education to build a solid knowledge base, rather than just focusing on tool usage.

Podcast:

About:
This study examines the state of digital transformation (DT) within the craft industry, challenging the
common perception that it lags behind other sectors. Through a systematic literature review of 141
academic and industry sources, the paper maps the application and influence of specific digital
technologies across various craft sectors. The research categorizes these technologies to
understand their impact on value creation, value proposition, and customer interaction.

Problem:
The craft and skilled trades industry is often viewed as traditional and slow to adopt new
technologies, which can lead to missed business opportunities and a misunderstanding of its
innovative potential. There is a lack of detailed, sector-specific knowledge on how different craft
businesses are actually implementing digital tools. This study addresses this gap by clarifying the
specific patterns of digital adoption and identifying areas for future growth and research.

Outcomes:
- Contrary to popular belief, the craft industry shows active and diverse patterns of digital
technology adoption, which vary significantly by sector.
- Craft businesses are not resistant to technology but are strategically selective, often adopting
tools to enhance efficiency and production while preserving core artisanal values.
- The primary focus of technology adoption is on 'value creation' (optimizing processes), with less
emphasis on 'customer interaction' or 'value proposition' (product innovation).
- Sectors like construction and textiles are leveraging advanced technologies like AI, IoT, and BIM,
while traditional crafts focus more on e-commerce and social media for marketing.
- The findings suggest that a one-size-fits-all approach to digital transformation is ineffective for the
craft industry, highlighting the need for context-specific strategies.

Podcast:

About:
This study investigates how using Generative AI (GenAI) impacts the quality of online consumer
reviews. Through an online experiment, researchers compared reviews written with a standard
template to those written with a template that integrated GenAI, measuring the writer's effort and
the informativeness of the final review.

Problem:
Writing detailed and helpful online reviews is a mentally demanding task for consumers, which can
result in lower-quality feedback. This research addresses how new AI tools can be leveraged to
overcome this challenge and improve the overall usefulness of user-generated content for both
businesses and other consumers.

Outcomes:
- Using Generative AI to assist in writing reviews significantly reduces the mental effort (cognitive
load) for the writer.
- Reviews written with AI help are more informative, covering a greater number of specific aspects
and topics about the product or service.
- AI-assisted reviews tend to have a more positive sentiment in the text, even when the associated
star rating is similar to manually written reviews.
- These reviews are also more linguistically complex, which could make them harder for the
average reader to understand, potentially limiting their practical value despite being more detailed.

Podcast:

About:
This study investigates the financial impact of digital transformation (DT) investments within the
European consumer goods industry. Using an event study for short-term effects and panel data
analysis for medium-term outcomes, the research differentiates between the market reaction to
single DT announcements and the value of sustained digital maturity. The goal is to provide
strategic insights into how companies can prioritize DT investments to achieve a competitive
advantage.

Problem:
Businesses are investing heavily in digital transformation, but it's often unclear if these expensive
initiatives provide a tangible financial return. Leaders struggle to justify investments with delayed or
intangible benefits, and it is unknown whether the market values one-off projects or a long-term
strategic commitment. This study addresses the gap by examining what kind of DT investment
strategy actually pays off in financial markets.

Outcomes:
- Single announcements of digital transformation investments do not create significant positive
short-term stock returns for a company.
- Sustained, long-term digital maturity is strongly correlated with higher company valuations
(market-to-book ratio) over the medium term.
- Companies with higher digital maturity benefit from lower costs of capital, as they are perceived as
less risky by both equity investors and debt providers.
- The financial rewards of digital transformation come from a consistent, cumulative strategy rather
than isolated, individual projects.
- A long-term commitment to digital capabilities signals a sustainable competitive advantage to the
market, leading to tangible financial benefits.

Podcast:

About:
This study investigates how employees' personality traits, sense of identity, and organizational
culture influence their adoption and use of generative AI (GenAI) in the workplace. Based on 23
expert interviews, the research identifies key behavioral patterns and proposes a framework for
understanding different employee responses to GenAI. The analysis reveals four distinct AI user
archetypes, highlighting the complex interplay between individual psychology and corporate
environment.

Problem:
As organizations rapidly integrate generative AI, they observe a wide range of employee reactions,
from transparent adoption to strategic concealment. This inconsistency creates challenges for
managing security risks, ensuring ethical use, and developing effective training programs. This
study addresses the gap in understanding the personal and cultural drivers behind why employees
choose to use AI openly, hide it, or resist it.

Outcomes:
- The study identified four distinct employee archetypes regarding AI use: 'Innovative Pioneers'
(open and identify with AI), 'Hidden Users' (identify with AI but conceal its use), 'Transparent Users'
(use AI openly as a tool), and 'Critical Skeptics' (cautious and avoidant).
- Individual personality traits, particularly negative ones like narcissism, can drive employees to hide
their AI use to maintain a competitive edge or create a false impression of expertise.
- Organizational culture is critical; open and innovative environments promote transparent AI
adoption, whereas rigid, hierarchical, or highly competitive cultures encourage concealment.
- Hidden AI usage, or 'Shadow AI', introduces significant risks, including data privacy violations,
security vulnerabilities, and the development of hidden skill gaps in the workforce.
- Companies should foster supportive cultures with clear ethical guidelines and targeted training to
encourage responsible AI integration and mitigate potential risks.

Podcast:

About:
This study introduces an advanced method for analyzing auction data to understand bidders' true
valuations. It extends a classic framework by replacing traditional statistical estimators with a novel
approach that combines optimal transport theory and equilibrium learning. This new technique aims
to provide more accurate and robust estimations of bidder behavior from observed bids.

Problem:
Auction designers need to understand bidders' private valuations to set optimal rules, such as
reserve prices, and maximize revenue. However, this information is hidden, and existing statistical
methods used to infer it from bid data are often unreliable, especially with limited data, leading to
biased estimates and poor strategic decisions.

Outcomes:
- The proposed method, based on optimal transport, more accurately estimates bidders' private
valuations from auction data compared to traditional techniques.
- This superior accuracy leads to better business decisions, such as setting optimal reserve prices
that can significantly increase an auctioneer's revenue, whereas older methods failed in the
experiments.
- The new framework is more robust and flexible, performing well even with limited or high-variance
data, making it more practical for real-world auction design.

Podcast:

About:
This study analyzes the scheduling of large maintenance vehicles for Germany's railway network,
using real-world data from Deutsche Bahn. It employs optimization models, including a greedy
heuristic and Mixed Integer Programming (MIP), to evaluate how factors like fleet size and
scheduling flexibility impact maintenance efficiency.

Problem:
Railway maintenance is essential but frequently causes major disruptions, delays, and capacity
reductions for both passenger and freight services. This inefficiency reduces the reliability and
attractiveness of the rail system, creating a need for better planning and resource allocation to
minimize service interruptions.

Outcomes:
- Organizational flexibility is more critical than fleet size; simply adding more maintenance vehicles
provides little to no benefit for completing maintenance tasks.
- The primary bottleneck is the availability of pre-defined maintenance time slots ('containers').
Allowing these slots to be reused multiple times can increase task completion to nearly 100%.
- Increasing the travel range for vehicles from their depots and extending shift lengths can improve
outcomes, but the benefits are less significant and come with practical trade-offs like increased
labor costs.
- Optimizing the size and reusability of maintenance slots is a more cost-effective strategy for
improving railway maintenance efficiency than investing in more physical assets.

Podcast:

About:
This paper conducts a comprehensive review of existing research on 'boundary resources'—the
tools like APIs and SDKs that connect digital platforms with third-party developers. By analyzing 89
academic publications, the study maps the current state of knowledge, identifies significant gaps,
and proposes a clear agenda for future research.

Problem:
Digital platforms are central to modern business, but how they manage third-party innovation is
poorly understood. This study addresses key knowledge gaps, including the lack of clarity on the
financial impact of opening a platform, an overemphasis on consumer-facing platforms (B2C) while
ignoring business-to-business (B2B) contexts, and the emerging role of AI as a developer.

Outcomes:
- Platform strategies are fragmented; companies focus on using boundary resources for control and
co-innovation but often miss their direct financial and strategic value.
- Most research focuses on consumer platforms (like app stores), meaning the findings may not
apply to business-to-business (B2B) ecosystems, which operate differently.
- There's a critical lack of a clear definition of what distinguishes a 'platform' from other software,
complicating strategic decisions about openness and control.
- The rise of AI and automation is changing who or what builds on platforms, requiring new
governance models to manage both human and machine developers (complementors).

Podcast:

About:
This study analyzes user behavior on a blockchain-based gambling platform called YOLO, which is
implemented as a smart contract. By examining over 22,800 gambling rounds from 3,306 unique
users, the research uses statistical models to understand how cognitive biases affect gambling
habits in a decentralized environment. The goal is to provide insights for regulators and user
protection initiatives.

Problem:
Traditional online gambling platforms pose societal risks and often evade regulatory oversight. The
rise of decentralized, blockchain-based gambling platforms worsens these issues by operating
without intermediaries, which makes enforcing user protection and legal compliance extremely
difficult. This study addresses the lack of understanding of user behavior and the unique risks
associated with these new platforms.

Outcomes:
- Cognitive biases, such as the 'anchoring effect' and 'gambler's fallacy', strongly influence users to
continue gambling, often leading to cycles of financial loss.
- The study found that users who repeatedly bet the same amount (anchoring) or increase their bets
after losing streaks are more likely to keep playing.
- The decentralized and pseudonymous nature of blockchain gambling makes it nearly impossible
to implement standard responsible gaming measures like deposit limits, age verification, or
self-exclusion tools.
- The findings highlight an urgent need for new regulatory frameworks tailored to decentralized
platforms, such as on-chain monitoring to detect risky behavior and the potential for blacklisting
gambling smart contracts.

Podcast:

About:
This study examines how to best integrate generative AI (genAI) tools on peer-to-peer (P2P) rental
platforms like Airbnb to help users write property listings. Using a controlled experiment with 244
participants, researchers tested how the timing (before, during, or after writing) and interactivity
(automatic vs. user-prompted) of AI suggestions influence user reliance on the technology.

Problem:
P2P rental platforms depend on unique, host-written descriptions to attract customers, but creating
this content can be time-consuming for hosts. While generative AI offers a solution to streamline
content creation, there is little guidance on how to design these AI assistants to maximize user
adoption and ensure they are genuinely helpful.

Outcomes:
- Offering AI writing assistance early in the content creation process significantly increases the
likelihood that users will rely on it.
- Users are less likely to adopt AI suggestions if they are introduced after they have already
invested effort in writing their own description.
- Allowing users to actively prompt the AI for suggestions has a small, but positive, effect on their
reliance compared to receiving automatic suggestions.
- When users feel mentally overloaded by a task, they are less likely to trust and use AI-generated
suggestions.
- To maximize adoption, platforms should integrate AI writing tools that engage users at the very
beginning of the listing process.

Podcast:

About:
This study analyzes why research on human trust in AI yields conflicting results, leading to both
over-trust and under-trust. Through a systematic literature review of 59 empirical studies, the paper
develops a new framework to explain these inconsistencies. The framework emphasizes that the
specific context in which a decision is made is crucial for understanding how people trust and rely
on AI systems.

Problem:
For humans and AI to collaborate effectively on complex decisions, humans must have an
appropriate level of trust in the AI. However, people often misjudge AI capabilities, leading to either
blind over-reliance on flawed advice or dismissal of correct advice. This research addresses the
gap in understanding why this happens by showing that previous studies have often overlooked key
contextual factors related to the human user, the AI system, and the decision itself.

Outcomes:
- Trust in AI is not static but is heavily shaped by the specific context of the interaction.
- Key AI-related factors influencing trust include its performance, explainability, adaptability, and
how human-like it appears.
- Key decision-related factors include the level of risk, complexity, morality, and objectivity of the
task.
- Key human-related factors include the user's expertise, familiarity with AI, cognitive biases, and
sense of control over the decision.
- By considering these factors together, businesses can better predict and manage user trust to
improve the outcomes of human-AI collaboration.

Podcast:

About:
This study investigates the reasons behind the low adoption rate of blockchain technology in the
German pig value chain, despite its widely recognized potential benefits for food supply chains.
Through semi-structured interviews with eight industry experts, the research identifies the key
factors that prevent the implementation of blockchain in a real-world setting.

Problem:
Blockchain technology is often presented as a solution to enhance transparency, traceability, and
trust in food supply chains. However, there is a significant gap between the technology's perceived
benefits and its actual adoption, as seen in the German pig industry, a major player in the European
market. This study addresses why this promising technology is not being implemented in practice.

Outcomes:
- Stakeholders believe their current systems for data exchange and management are sufficient and
see no significant additional benefit from implementing blockchain.
- Trust, a key feature offered by blockchain, is already well-established within the industry through
long-standing business relationships and organizational ownership structures, making the
technology's trust-building aspect redundant.
- There is no strong market demand from within the value chain or from end consumers for the
specific features blockchain offers, such as enhanced traceability or immutability.
- Significant barriers, including the high cost of investment, lack of financial slack, and missing
digital infrastructure, outweigh the technology's potential advantages.
- The decision not to adopt blockchain is a rational choice based on a cost-benefit analysis where
existing solutions and trust mechanisms are deemed adequate for current needs.

Podcast:

About:
This study explores how organizations use storytelling, or narratives, to improve the success of
their digital transformation efforts. Through a series of expert interviews, the research analyzes the
key functions and characteristics of effective narratives in managing the human side of
technological change.

Problem:
Many digital transformation initiatives fail due to human-related factors, particularly stakeholder
resistance. While narratives are often used to shape these efforts, there is a lack of understanding
among both researchers and practitioners about how to leverage them effectively to mitigate risks
and ensure successful outcomes.

Outcomes:
- Narratives serve a key motivational and transformative function, helping people make sense of
change and giving purpose to transformation efforts.
- Well-crafted narratives can reduce stakeholder resistance and actively engage employees in the
transformation process.
- The context and specific 'boundary conditions' are important for a narrative to be successful.
- An overarching narrative is essential for defining and guiding digital transformation initiatives,
especially those involving AI.

Podcast:

About:
This study investigates how workers in traditional, non-platform organizations perceive algorithmic
control (AC) systems that manage their tasks. Using a comparative analysis of 92 logistics workers,
the research explores how individual character traits, specifically competitiveness, influence
whether employees view AC as legitimate in terms of fairness, autonomy, and skill development.

Problem:
As traditional companies increasingly use algorithms to manage employees, they lack
understanding of how different types of workers will perceive these systems. This knowledge gap is
critical because a system that motivates one employee might be rejected by another, impacting
productivity, morale, and the overall success of technological implementation.

Outcomes:
- Employee personality is a major factor in the acceptance of algorithmic management; highly
competitive workers view it far more positively than their non-competitive colleagues.
- Competitive employees see algorithmic rating and recommendations as fair and beneficial for their
professional development, viewing it as a tool for performance maximization.
- Non-competitive employees tend to reject most forms of algorithmic control, finding them unfair
and detrimental to their skill growth.
- A 'one-size-fits-all' approach to algorithmic management is ineffective; businesses must design
flexible systems that cater to different worker personalities or risk alienating segments of their
workforce.
- Without considering personality, companies risk demotivating non-competitive workers, which can
lead to resistance, decreased engagement, and higher turnover.

Podcast:

About:
This study systematically reviews 64 empirical papers published between 2014 and 2024 to
understand what makes Digital Business Simulation Games (DBSGs) effective training tools. The
research identifies four key types of training outcomes—attitudinal, motivational, behavioral, and
cognitive. Based on this evidence, the paper develops a practical framework of design guidelines
for creating and implementing impactful business simulations.

Problem:
Companies and universities increasingly use Digital Business Simulation Games to teach complex
decision-making in a risk-free environment. However, research on their effectiveness is fragmented,
and there is a lack of clear, evidence-based guidelines, leading to inconsistent training results and
an inability to maximize the impact of these powerful learning tools.

Outcomes:
- To improve user satisfaction, design games with realistic business scenarios, high-quality in-game
information, and immersive visual and interactive elements.
- Boost trainee motivation by incorporating compelling storylines, progressively increasing decision
complexity, and adding competitive elements like leaderboards.
- Enhance critical skills like teamwork and problem-solving by designing games that actively
encourage participant interaction and collaboration.
- Maximize learning by aligning the game's difficulty with participant skill levels and providing
continuous feedback through reports, hints, and tutorials.
- Successful implementation requires well-trained instructors, clear pre-training briefings to set
expectations, and guided debriefing sessions to connect the simulation experience with real-world
application.

Podcast:

About:
This study details the development of an AI system to detect online hate speech, particularly in
multimodal formats like memes. The system uses a novel approach where multiple specialized AI
'agents' engage in a court-inspired debate to classify content. This method is designed to make the
AI's decision-making process transparent and understandable, incorporating a 'human-in-the-loop'
for user oversight and refinement.

Problem:
Online hate speech is a widespread problem, but automated AI detection systems are often like
'black boxes,' making it difficult to understand why they make certain decisions. This lack of
transparency undermines trust and complicates error analysis, creating a significant obstacle for its
use by social media companies and legal professionals who require clear justifications for content
moderation.

Outcomes:
- The proposed multi-agent system (MAS) achieved 70% accuracy in detecting hateful memes,
outperforming a standard single-agent AI model across all tested metrics.
- The system's unique debate-based framework successfully enhances explainability, allowing
users to follow the reasoning behind a classification.
- By externalizing the decision logic into a structured debate between agents, the model provides
transparent and auditable insights for moderators and legal professionals.
- The inclusion of a human-in-the-loop workflow enables users to review, refine, and customize the
system's rules, increasing reliability and trust.

Podcast:

About:
This study investigates the design of speech-based assistance systems (SBAS) for automating the
task of taking meeting minutes. Researchers developed and evaluated a prototype with different
levels of automation to understand how to balance the benefits of efficiency with potential
drawbacks for the user.

Problem:
While AI tools that automate tasks like minute-taking can offer significant economic benefits and
efficiency gains, they can also have negative consequences. High levels of automation risk
reducing employee satisfaction, their sense of professional identity, and their identification with their
work.

Outcomes:
- A higher level of automation successfully improves the capture and processing of information from
meetings and reduces the user's cognitive effort.
- However, high automation can decrease user satisfaction and their sense of ownership over the
final work. Users reported greater satisfaction with partial automation systems where they remained
more involved.
- Highly automated systems can lead to 'cognitive complacency,' where users accept
system-generated content without proper review, increasing the risk of errors or irrelevant
information being included.
- To be effective, assistance systems should be designed to augment human work, not just replace
it. The key is to balance automation with features that ensure meaningful user integration and
oversight.

Podcast:

About:
This study introduces a novel two-stage clustering method to improve the accuracy and
interpretability of house price predictions. The approach first groups German properties by location
and then further segments them based on property characteristics, applying interpretable machine
learning models to each resulting cluster. The methodology is tested on a dataset of over 43,000
German house property listings from 2023.

Problem:
Predicting house prices accurately is challenging due to significant variations across local markets.
Current valuation methods force a trade-off between simple, transparent models that are often
inaccurate, and complex 'black-box' models that are accurate but lack explainability. This study
addresses the need for a method that is both highly accurate and easily understandable for
real-world application.

Outcomes:
- The two-stage clustering approach dramatically improves house price prediction accuracy,
reducing the mean absolute error by up to 58% for Linear Regression and 36% for Generalized
Additive Models (GAMs) compared to models without clustering.
- The model reveals that the influence of property features, such as construction year and living
space, varies significantly across different market segments (clusters).
- This method provides interpretable, actionable insights into local price drivers, offering a clear
advantage over opaque 'black-box' models.
- The findings confirm that geographical location is the most critical factor in determining property
values, justifying its use as the initial step for market segmentation.

Podcast:

About:
This study explores the key factors driving women's engagement with digital fitness apps by
analyzing over 34,000 user reviews. Using computational topic modeling, the research maps user
concerns to the core psychological needs of Self-Determination Theory—autonomy, competence,
and relatedness—to create a framework for designing more effective fitness technologies.

Problem:
Many digital fitness apps fail to maintain long-term user adherence because they often lack a
theoretical foundation and adopt a 'one-size-fits-all' approach. This research addresses the specific
need to understand the psychological drivers for women, who are consistently less physically active
than men, to help create more motivating and effective health solutions.

Outcomes:
- Autonomy is the most critical factor for women using fitness apps; users desire control, flexibility,
and personalization, such as customizable plans and workouts adaptable to different environments.
- Competence is the second most important driver, where users seek apps that not only track
progress but also provide structured feedback and help them develop new skills.
- Social connection (relatedness) is also significant, with users valuing supportive coaches,
community interaction, and relatable content from digital influencers, particularly concerning
maternal health.

Podcast:

About:
This research paper conducts a systematic literature review to categorize the various interfaces used for interacting with Data Trusts, which are intermediaries for fiduciary data sharing. The study classifies interfaces into two primary groups: human-system interfaces for user interaction (e.g., dashboards) and system-system interfaces for automated machine communication (e.g., APIs). The goal is to create a foundational understanding to guide the practical implementation of trustworthy data sharing ecosystems.

Problem:
Data Trusts are emerging as a vital mechanism for secure and ethical data sharing, but their successful implementation is hindered by a lack of clear standards and understanding of how different actors should interact with them. There is a significant research gap concerning the specific design and systematization of user and technical interfaces. This ambiguity makes it difficult to build robust, interoperable, and user-friendly Data Trusts that can be widely adopted.

Outcomes:
- Interfaces for Data Trusts are categorized into two essential types: Human-System (e.g., Graphical User Interfaces, dashboards) and System-System (e.g., APIs, messaging interfaces).
- The study reveals that existing research lacks clear, detailed specifications and rigorous analysis of how these interfaces are implemented in Data Trusts.
- There is a critical need for standardized, interoperable technical interfaces (APIs) to ensure seamless data exchange between different systems and data sources.
- The research highlights a deficiency in literature regarding the connection between interfaces, different data structures (structured, semi-structured, unstructured), and data storage systems.
- The paper provides a comprehensive overview that serves as a foundational step for future development and practical implementation of interfaces in Data Trust ecosystems.

Podcast:

About:
This study presents the conceptual design and evaluation of the "PV Solution Guide," a user-centric prototype for a photovoltaic (PV) system decision support system. Researchers used qualitative interviews to identify homeowner needs, then developed an interactive prototype designed to provide personalized guidance. The prototype's usability and trustworthiness were then tested and compared against an established, state-of-the-art government tool.

Problem:
Homeowners considering solar energy often face significant hurdles due to rigid and impersonal online consultation guides. These tools frequently fail to address unique home characteristics or varying levels of user knowledge, leading to a lack of trust and decision paralysis. This information gap creates a major barrier to the adoption of renewable energy solutions.

Outcomes:
- The "PV Solution Guide" prototype significantly outperformed an existing government tool in both usability and user trust.
- Users found the prototype much easier to use, with a usability score of 80.21 compared to the existing tool's 56.04.
- The prototype was perceived as more trustworthy, particularly in its competence (ability to perform its function) and benevolence (acting in the user's best interest).
- Key features driving success were its ability to adapt to the user's knowledge level and specific house type, transparent cost breakdowns, and an interactive 3D model.
- A dynamic, conversational, and personalized approach is more effective for guiding consumers than static, one-size-fits-all online calculators.

Podcast:

About:
This study outlines the development of a framework for AI-driven meal demand prediction systems, specifically focusing on the airline industry. Using a Design Science Research approach, the authors conducted interviews with industry experts to identify system requirements. These requirements were then used to create nine core design principles and a feasible system architecture, which was validated through a prototype.

Problem:
Airlines and catering services struggle with accurately forecasting in-flight meal demand, which leads to significant food waste, financial losses, and potential customer dissatisfaction if meals are unavailable. Traditional forecasting methods are often inadequate, and despite the potential of Artificial Intelligence (AI) to improve accuracy and efficiency, its adoption in the industry remains extremely low.

Outcomes:
- The research identified nine essential design principles for building effective AI-driven meal demand prediction systems, such as Data-Driven Forecasting, Automated Data Integration, and Sustainable & Waste-Minimising Design.
- A successful system must integrate various data sources, including historical internal data and external factors like weather and holidays, to improve prediction accuracy.
- A proposed system architecture demonstrates how these principles can be implemented through a central database, a machine learning model, and APIs for integration with existing airline systems.
- Key success factors include providing a transparent and user-friendly interface for catering managers, ensuring data privacy, and balancing cost-efficiency with customer satisfaction and sustainability goals.
- While focused on airlines, the design principles are broadly applicable to other industries dealing with perishable goods and complex demand forecasting.

Podcast:

About:
This study analyzed approximately 28,000 speeches from the German parliament (the Bundestag) over the last five years using artificial intelligence. The researchers developed machine learning models to automatically classify the topic and the sentiment (i.e., the critical or supportive tone) of each speech to uncover large-scale patterns in political communication.

Problem:
Amid growing public distrust in political institutions, it is often difficult for outsiders to get an objective understanding of political priorities and strategies directly from parliamentary debates. This study addresses this gap by providing a data-driven tool to analyze political discourse, offering transparent insights into how parties frame issues and change their communication styles based on their political roles.

Outcomes:
- A party's role as government or opposition is the strongest predictor of its communication style; opposition parties consistently use far more critical language.
- When parties switch roles after an election (e.g., moving from government to opposition), their tone changes dramatically, becoming significantly more negative.
- Parliamentary debates are dominated by three main topics: Economy & Finance, Social Affairs & Education, and Foreign & Security Policy, which together account for over 70% of discussions.
- Parties at the political extremes use a consistently high level of negative rhetoric, with over 80-90% of their speeches having a critical tone.
- Parties are often most critical when discussing their own core policy areas, strategically using negative sentiment to emphasize the urgency of their proposed solutions.

Podcast:

About:
This study investigates the unique challenges faced by Artificial Intelligence (AI) startups, including data acquisition, talent recruitment, regulatory hurdles, and competition. Using a qualitative approach based on ten semi-structured interviews with founders, the research applies effectuation theory to identify practical strategies for navigating the dynamic and uncertain AI business environment.

Problem:
While investment in AI startups is surging, these companies face distinct obstacles not seen in traditional tech ventures. Current business literature often fails to differentiate AI startups from other digital companies, leaving a gap in understanding how founders can effectively manage high uncertainty, data dependency, and a rapidly evolving regulatory landscape.

Outcomes:
- AI startups face key challenges in securing high-quality data, affordable AI models, and skilled talent. Successful founders mitigate these issues by partnering with customers for data, using open-source tools for lean development, and leveraging university networks for hiring.
- Instead of direct competition, effective startups differentiate by tailoring solutions for specific customer needs and building deep domain relationships, creating a competitive advantage that is difficult to replicate.
- Regulatory hurdles, such as the EU AI Act, can be transformed from a burden into a strategic asset. Proactive compliance can build trust with enterprise customers and create a barrier to entry for competitors.
- Founders successfully navigate market uncertainty not by predicting the future, but by acting adaptively. They leverage existing resources, form strategic partnerships, focus on affordable risks, and iterate based on customer feedback.

Podcast:

About:
This study investigates how to design enterprise systems (ES) that can meet the needs of fast-growing digital startups. Using a design science research approach, the authors interviewed representatives from 11 startups to identify key requirements for scalable business processes. The research resulted in nine specific design principles for creating flexible, adaptable, and scalable enterprise systems suitable for dynamic environments.

Problem:
Traditional enterprise systems are typically rigid and assume stable business processes, which makes them unsuitable for dynamic organizations like digital startups. These fast-scaling companies require flexible, adaptable systems to support continuous change and growth, but there is limited guidance on how to design ES that meet these specific demands. This study addresses this gap by exploring the unique ES needs of startups to propose a more suitable architectural approach.

Outcomes:
- The research identified nine core design principles for scalable enterprise systems tailored to startups: efficiency through automation, integration, data-driven decision-making, archiving, flexibility, scalability, performance, transparency, and user-centered design.
- A modular system architecture is proposed, featuring a central workflow engine, an extensible API, and integrated data services to support rapid adaptation and growth.
- Unlike traditional systems that prioritize rigid reliability, the proposed ES architecture emphasizes transparency and traceability through logging, which enables safe experimentation and continuous improvement.
- The findings highlight a critical need for systems that can seamlessly integrate with existing tools to reduce data redundancy, minimize manual effort, and support agile workflows.
- The study concludes that for startups, enterprise systems must be designed with process adaptability as a core feature, not an exception, to support their evolving business needs.

Podcast:

About:
This study investigates how AI-powered Intelligent Personal Assistants (IPAs) contribute to process innovation and improve business workflows. Using the Task-Technology Fit theory, the researchers analyzed data from interviews with twelve professionals to understand how these tools are integrated, their benefits, and their limitations in a real-world work context.

Problem:
Businesses are rapidly adopting AI, but there is limited understanding of how employees practically use tools like ChatGPT to innovate their work processes. This research addresses the gap by exploring the specific ways IPAs are integrated into daily workflows and how they enhance adaptability beyond simple task automation.

Outcomes:
- AI assistants enhance work processes in four key areas: guidance and problem-solving, decision support and brainstorming, workflow automation, and communication support.
- Key drivers for adopting these tools include social influence (word-of-mouth), the need for greater efficiency, and professional curiosity.
- Adoption is currently limited by challenges such as poor integration with existing software, data privacy concerns, and the AIs' memory and reasoning constraints.
- Regular users leverage IPAs for strategic and creative tasks, while occasional users focus on simple, repetitive work like documentation.
- Future improvements desired by users include better workflow integration, personalized responses, and enhanced security and ethical features to reduce bias.

Podcast:

About:
This study investigates how Artificial Intelligence (AI) creates business value specifically for platform-based companies. Using a two-part, mixed-methods approach, the research first develops a framework to classify AI applications based on expert interviews and then proposes an empirical study to measure their impact on shareholder value. The goal is to clarify how different AI tools augment organizational capabilities and contribute to a firm's bottom line.

Problem:
Businesses are investing billions in AI, but many projects fail to deliver significant returns. Existing research often examines AI applications in isolated contexts or too broadly, which doesn't help managers understand how different types of AI strategically contribute to business value. This gap in knowledge makes it difficult for companies to effectively deploy AI to improve processes, enhance services, and gain a competitive edge.

Outcomes:
- The research identifies 14 distinct AI applications used by businesses, categorizing them into four main groups: process automation, process augmentation, product/service offerings, and ecosystem development.
- AI creates value internally by automating tasks, improving efficiency, and enhancing internal knowledge sharing.
- AI creates value externally by enabling new products and services, improving personalization, and orchestrating interactions within the business ecosystem.
- The study provides a clear framework linking specific AI applications to their impact on either internal operations or external market offerings, helping to guide strategic investment.

Podcast:

About:
This study provides a comprehensive overview of Generative AI (GenAI) applications for business process optimization. Through a systematic literature review and qualitative content analysis, the paper assesses the maturity level of various GenAI use cases across different segments of the corporate value chain.

Problem:
While the potential of Generative AI is enormous, its novelty makes it difficult for businesses to identify where the technology is mature enough for effective implementation. This research addresses this gap by systematically mapping which business areas can already benefit from GenAI and which require further development, providing a clear guide for managers and researchers.

Outcomes:
- The most mature and well-researched applications of Generative AI are in technical domains, particularly within the production sector.
- Product development and maintenance & servicing are the business processes with the most advanced and successfully implemented GenAI solutions.
- Applications in areas with high levels of human interaction, such as marketing and sales, are currently less developed and represent emerging fields for research and implementation.
- The production segment of the value chain contains the highest number of mature GenAI use cases compared to other corporate functions.

Podcast:

About:
This study develops a new method, called RE-SIDE, for designing innovative services within complex business ecosystems that are enabled by shared data platforms. Using a design science research approach, the paper extends an existing industry standard to better account for the numerous actors and dynamic changes in these data-driven environments. The method's effectiveness is demonstrated through a real-world case study involving a "Culture Wallet" service.

Problem:
Traditional methods for creating new services are designed for simple, one-on-one interactions between a provider and a customer. These approaches are insufficient for modern digital ecosystems where numerous independent organizations must collaborate and share data to create value. This leaves a critical gap for a framework that can guide innovation in these complex, multi-actor settings.

Outcomes:
- The research introduces the RE-SIDE method, a practical framework for designing services in multi-partner, data-rich business ecosystems.
- It enhances existing service design standards by adding two crucial new phases: 'ecosystem analysis' and 'ecosystem transformation'.
- The 'ecosystem analysis' phase helps a company understand its position, identify potential partners, and assess opportunities and risks within the broader market landscape.
- The 'ecosystem transformation' phase provides a process for continuously monitoring and adapting services to ongoing changes in the ecosystem, ensuring long-term relevance.
- The method's application is proven through the design of a "Culture Wallet," showing how cultural institutions can collaborate using shared data to create new value for customers.

Podcast:

About:
This study proposes an online intervention method called "prebunking" to counter the spread of anti-democratic attitudes on social media. Based on the concept of attitude inoculation, the intervention aims to build a "future-oriented mindset" in individuals, thereby strengthening their mental resistance to manipulative content. The effectiveness of this approach will be tested through a controlled, randomized online experiment.

Problem:
Anti-democratic content on social media poses a significant threat to the stability of Western societies by increasing political cynicism and decreasing support for democratic institutions. Current countermeasures from tech companies and governments are often insufficient, and some interventions have even backfired. This research addresses the gap for a user-centered psychological approach that can proactively "inoculate" individuals against harmful persuasion.

Outcomes:
- Proposes a novel 'prebunking' intervention designed to build individuals' cognitive immunity against anti-democratic content online.
- The intervention aims to foster a 'future-oriented mindset', which is theorized to act as a protective factor against adopting anti-social and anti-democratic views.
- The study will experimentally test if this intervention can reduce pre-existing anti-democratic attitudes and lower the likelihood that individuals will share such content.
- The research seeks to establish a foundation for using psychologically-grounded interventions to safeguard individuals from manipulative online influences.

Podcast:

About:
This study introduces BPMN4CAI, a standard-compliant extension for Business Process Model and Notation (BPMN) designed to integrate dynamic Conversational AI into business workflows. Using a Design Science Research methodology, the paper develops and evaluates an approach that systematically extends existing BPMN elements with specialized components. The framework's applicability is demonstrated through a practical case study.

Problem:
Standard Business Process Model and Notation (BPMN) is designed for predictable, deterministic processes, making it ill-suited for modeling modern Conversational AI systems like chatbots. These AI systems exhibit dynamic, context-sensitive, and non-deterministic behaviors that are difficult to represent, creating a practical and methodological gap for integrating them transparently into formal business process designs.

Outcomes:
- Developed BPMN4CAI, a standard-compliant extension that allows for the formal modeling of Conversational AI within business processes.
- The extension introduces new elements like the 'Conversational Task,' 'AI Decision Gateway,' and 'Human Escalation Event' to represent dynamic dialogue, AI-based decisions, and handovers to human agents.
- The results show that BPMN4CAI improves the ability to model adaptive decision-making, manage conversation context, and create transparent AI interactions.
- A proof-of-concept in the insurance industry confirmed the practical applicability and benefits of the extension over traditional BPMN.
- The study also identifies limitations, noting that modeling highly complex, non-deterministic process paths remains a challenge.

Podcast: