We study the risk of collusive behavior when using Reinforcement
Learning (RL) algorithms to decide on pricing strategies in competitive markets.
Prior research in this field focused exclusively on Tabular Q-learning (TQL) and
led to opposing views on whether learning-based algorithms can lead to supra-
competitive prices. Firms are increasingly using Deep Reinforcement Learning
(DRL). Our work contributes to this ongoing discussion by providing a more
nuanced numerical study that goes beyond TQL by additionally capturing off- and
on-policy based DRL algorithms: Deep Q-Networks (DQN) and Proximal Policy
Optimization (PPO). We study the dynamics of these algorithms in a Bertrand
competition and show that algorithmic collusion indeed depends on the algorithm
used. In our experiments TQL exhibits higher collusion and price dispersion phe-
nomena, while DQN and PPO show lower collusion tendencies, with PPO, in
particular, achieving a large proportion of competitive outcomes. We further show
that algorithm dynamics are sensitive to general algorithm design decisions.

The rapid evolution of the educational landscape, accentuated by the rise of Generative Artificial Intelligence (GAI), calls for a re-evaluation of digital education methodologies to harness new technologies and meet the changing needs of learners. Focusing on the potential of GAI, this study introduces a novel approach called generative feedback and explores its impact on the motivation, learning experience and performance of students in higher education. Therefore, this study presents a prototype that implements a GAI-driven chat using OpenAI’s GPT-4 turbo model to provide feedback to students. Using an overlapping study design that combines a longitudinal field study and a two-stage online experiment, this study investigates how generative feedback affects learners. Initial findings suggest that GAI-driven chatbots can provide meaningful feedback to students and enhance their learning experience, setting the stage for further investigation towards a better theoretical understanding of the design and practical application of GAI-driven chatbot feedback.

In human-computer interaction (HCI) research, ensuring human well-being is seen as a key challenge. The consideration of well-being orientations, e.g. hedonia (aiming for pleasure, comfort, and relaxation) and eudaimonia (striving for authenticity, meaningfulness, excellence, and growth) is still at an early stage with regard to identification, measurement, and differentiability concerning expectations of technology. Particularly in the architectural design of recommender systems (RS), work is yet in the fledgling stage on how well-being can be considered in the development of RS. In this research in progress, we aim to present initial findings based on our labeled dataset (using the HEMA-Revised (HEMA-R) Scale (hedonia and eudaimonia)) of 2,946 items (movies, songs, and books) with the help of 229 participants. With a first architectural design on how the idea of well-being-centered RS can be implemented, we expand the research on group RS by introducing well-being-centered group RS.

Today’s society is aging and is becoming more care-dependent. Due
to the increasing scarcity of professional care workers, we are witnessing the
growing importance of non-professional family care. As the demands for caring
can be challenging, digital health applications play a crucial role in assisting with
caregiving tasks. These apps help family members, friends, and neighbors who
are typically not trained in caregiving. While this is a promising class of tools,
there is still limited knowledge on how to design them effectively. In this re-
search-in-progress paper, we reviewed 29 scientific papers on requirements for
informal caregiving support applications (ICSA) and derived 23 design proposi-
tions. Our work raises awareness of this particular tool class, provides an over-
view of options for ICSA designers, and paves the ground for future research.

Enter the web application, type in a question, and get a human-like answer in no time. Especially with the advent of ChatGPT, text-generating artificial intelligence permeates daily life. As a result, end-users are trying out new applications bearing risks, such as overconfidence. This research-in-progress paper investigates the main factors affecting end-user perception regarding human-like AI-generated output and corresponding trust. With the overarching goal of appropriate protection by creating a standardized information structure for integration into websites as our artifact, we conduct a structured literature review in the first step to determine what causes overconfidence and the issues that need to be addressed by an appropriate solution. Therefore, we contribute to the broader aim of preventing end users from misinterpreting AI output. Our findings highlight AI literacy, difficulties in detecting misinformation, and a lack of transparency and explainability as critical factors to consider during solution development.

This working paper explores the concept of cultural digital twin cities,

proposing a novel approach to incorporating cultural and social data into the

digital modeling of urban environments. Recognizing the limitations of current

digital twins, which largely focus on physical and infrastructural elements, this

work in progress paper emphasizes the significance of uncurated data sources,

such as open street map imagery, social media conversations, and its metadata.

This approach not only enhances the understanding of urban life beyond tangible

aspects, but also includes currently inaccessible aspects of the city in urban

development. The paper sets a foundation for future research, development of

frameworks, and software dedicated to the realization of ‘other’ digital twin

cities, aiming to bridge the divide curated and uncurated urban dimensions.

Polarization occurs within society's networks when highly connected groups form with weak intergroup links, leading to echo chambers and filter bubbles. These phenomena hinder exposure to diverse viewpoints, posing significant challenges to democracy and societal welfare. Despite extensive research on measuring and mitigating social network polarization, the effectiveness of existing metrics remains largely uncharted. This study reevaluates these metrics and recommender system-based reduction strategies, pinpointing inherent limitations. It highlights key factors influencing polarization and adopts a design science research approach to craft a recommender system-based model for reducing polarization in online networks, recognizing its complex nature.

In this paper, we introduce the schema representation problem using the example of dining philosophers: The difference between a system model of five eating philosophers and a schema model for a set of eating philosophers is of major importance. In a Petri net model, each philosopher and each fork would be considered as separate entities with their relating states and transitions. However, this approach lacks due to scalability and dynamic behavior, as adding more philosophers and forks significantly increases the model’s size. To model any set of dining philosophers, a Petri net schema is useful. However, there is no modeling technique to model an infinite set of philosophers and forks, and to access its single elements. To address this problem, we provide the elm-notation, which allows us to dynamically unfold and aggregate any sets whereby behavior can be described for each philosopher and fork on schema level.

With the digitalization of healthcare and the wide distribution of smartphones, mobile health applications are increasingly established in medical care. Older adults, especially those affected by chronic conditions, can benefit from these applications but use them comparatively rarely. Studies on their adoption behavior are scarce. Through user tests, this study examines older adults’ adoption behavior when using a mobile health application prototype. The Senior Technology Acceptance Model serves as an expandable theoretical framework. Preliminary results reveal that, alongside prior technical experience, self-perception has the most decisive influence on actual usage behavior. Physicians play a crucial role in the adoption of mobile health applications, as they are trusted to assess their utility and necessity but are barely considered in technology acceptance models. Further research is required to substantiate these findings.

Mobility-on-Demand services will soon have the opportunity to integrate autonomous vehicles into their operations. During the transition towards full automation, MoD systems will operate a mixed fleet, where human-driven (HVs) and autonomous vehicles (AVs) coexist. In such scenarios, the AVs fully complies with the operator’s decisions while HVs must be incentivized through wages—i.e. the size and locational distribution of the human fleet is impacted by the wage decisions of the operator. In this paper, we present a data-driven information system to manage such mixed fleets. The system decides on trip prices, relocation of AVs, and wages of the HVs. Thereby, it takes into account the effect of wage decisions on the behavior of HVs. Validation of the system is conducted through a case study based on observed trip demand data in New York City. Our study demonstrates the importance of accounting for wage-dependent HVs' supply in this new scenario.

This study presents an innovative approach to advancing interoperability in information systems through the development of an Artificial Intelligence (AI)-based data mediator. Although standards have contributed to interoperability among disparate systems, the lack of universal standards still requires tools for data mediation. To reduce the substantial need for manual configuration of these systems, this paper outlines a strategy for translating data between two systems with different data schemas automatically. Unlike traditional methods, the proposed data mediator leverages recent advancements in AI to facilitate automatic mapping of heterogeneous data.

Cyberattacks and data breaches lead to high costs for organizations worldwide. Information security education and training awareness programs are one of the most important countermeasures. Here, assessing individuals' level of information security awareness is a crucial task. Regarding this, one of the major challenges is to measure security behavior, a core dimension of security awareness. This is because it is often assessed indirectly through questionnaires, which could bias metrics. Therefore, our overarching goal is to develop a more holistic metric that considers and integrates actual human behavior. In this design science research study, we present the status quo of our research, namely a prototypical instance for such a measurement approach, and initial meta-requirements based on two design iterations and pilot tests: a scavenger hunt to measure the consequences of real-world interactions, based on the Human-Aspect-of-Information-Security-Questionnaire as a scientific foundation.

In an era where digital transformation shapes most facets of business and society, haptic Design Thinking workshops emerge as a valuable driver to foster co-creation and human-centric design among interacting stakeholders. Still, a need persists to capture workshop results in a digital and machine-processable manner. Scene2Model exemplifies these capabilities by enabling an automated transformation of physical scenes into conceptual modeling-based representations. To support this process, novel features are introduced within this work that leverage generative Artificial Intelligence (AI), such as Large Language Model-based object description and attribute generation, along with comprehensive scene summaries. Moreover, the application of the Scene2Model tool in collective intelligence environments is supported by automated postings of results, thus exemplifying the potential to enhance collaborative evaluation and feedback. This approach emphasizes the potential of combining haptic Design Thinking with advanced AI technologies, marking a significant advancement in the refinement of Digital Twins for Haptic Design Thinking.

Despite its growing potential for value creation, German small and medium-sized enterprises (SMEs) face significant challenges in identifying strategies for adopting artificial intelligence (AI) in their operations. To address this important issue in a comprehensive practical guide, we follow a design science approach to develop a novel, tailored methodology for improving AI integration. Consisting of two phases – success factor analysis and roadmap identification – our methodology aims to equip SMEs with knowledge for AI integration based on the TOE (Technology, Organization, Environment) framework. While we demonstrate the application of our methodology within a single case study and present a brief evaluation to assess its effectiveness, our research in progress leaves the systematic identification of areas to refine our methodology to future research. Nevertheless, our findings contribute to research and practice by providing an applicable, theory-based methodology specifically designed to guide German SMEs in identifying actionable steps toward AI adoption.

This study investigates the presence of stereotype bias in the first images produced by repeatedly prompting text-to-image systems, particularly texttt{DALL-E3}. The motivation for this study lies in the concern that the primacy effect could amplify stereotypical biases in users' perceptions. Through a questionnaire survey, participants were asked to rank images based on perceived stereotypical content. Results indicate a pronounced bias in the first images generated, with statistical analysis confirming a significant association between the order of image presentation and the perception of stereotypical bias.

Companies have pursued automation for years. Small and medium-sized enterprises (SMEs) are still far from achieving complete hyperautomation of their entire company or individual areas of their business. There is uncertainty surrounding the causes and strategies for achieving a fully networked and autonomous process chain. In a DSR process, we aim to develop a tool that assists companies in recognizing the opportunities of hyperautomation, advancing their company step by step, and identifying current weaknesses. First, we analyze the problem space through focus groups with German SME partners. We extend the knowledge base through structured literature analysis to develop a comprehensive overview of hyperautomation barriers. Building on this, we aim to establish a success model before developing a maturity model as a second artifact as well as a canvas tool for evaluation. In this way, we support companies on their way to hyperautomated business processes.

The tourism industry plays a crucial role in many countries and has suffered under the global events of the last years (e.g., COVID-19, economic recession) with a decline in visitors and economic contributions. Through digitalization efforts of cities and states, smart tourism destinations use emerging technologies like the Internet of Things and Smart Services to provide rich on-site experiences for visitors, hoping to increase their competitiveness and attract new tourists. By conducting a design science research project in collaboration with a German municipality, this paper aims to explore how digital assistants should be designed in the context of smart tourism. Based on expert interviews and a multi-case study of 37 tourism apps, we were able to develop twelve key design principles for smart tourism applications, evaluated through a prototypical instantiation, to create design knowledge for the digitalization of tourism.

This short paper addresses the complex relationship dynamics between a municipal website's gamified IS artifact design and the desired goal states of municipal trust. To better understand these relationship, a mixed-methods study is proposed that identifies relevant gamification features in a three-step process in Study 1 and then explores the relationship between gamificationand municipal trust using a survey in Study 2. The study results promise added value for (a) user-centered technology design and (b) novel feedback mechanisms to strengthen municipal trust. Additionally, we will contribute to existing gamifica-tion literature by identifying relevant gamification features in relation to the de-sign of municipal websites.

Agile Software Development Teams (ASDTs) employ agile methodologies to navigate the dynamic landscape of the software industry, responding effectively to constant change and innovation. The success of ASDTs relies significantly on the individuals’ mindset, referred to as the agile mindset (AM). Yet, a clear conceptualization of the AM remains elusive, as existing approaches do not meet scientific standards and fail to produce consistent results. This paper addresses this gap by systematically conceptualizing the AM within the context of ASDTs. Through a systematic review and evaluation of existing definitions and initial conceptualizations, we identify dimensions and consolidate insights, resulting in a comprehensive conceptualization of the AM. Our work represents a crucial step towards understanding the AM’s role in ASDTs and lays the foundation for our planned future work: the empirical validation and development of a standardized measurement instrument for the AM.

Transforming vehicle data into actual value propositions remains a challenging endeavor. Consequently, there is a growing recognition in the car industry that collaboration among various stakeholders is essential to leverage value from data, leading to the emergence of automotive data ecosystems. However, it remains unclear why actors participate in these ecosystems, especially when co-creating and realizing value from vehicle data is complex and challenging. Through a multi-case study involving 12 interviews, we provide preliminary insights into why actors engage in automotive data ecosystems. We contribute to the literature by illustrating how the context influences engagement in these ecosystems. We also add to the understanding of Original Equipment Manufacturer (OEM) dispositions, further unpacking the automotive data ecosystem and its actors.

Manufacturing companies face a vast increase of data. Connected sen-sors turn physically isolated objects into nodes in data communication networks. This development enables but also forces companies to harness their data to gain a competitive edge. In this regard, anomaly detection enables seamless processes, so that production failures can be avoided. Artificial intelligence (AI) and espe-cially machine learning and deep learning constitute instruments to leverage sta-tistical complexity necessary to identify anomalies in these vast amounts of data. AI-based anomaly detection has therefore been subject to an intensive academic discourse in Information Systems. This short paper provides preliminary results from a bibliometric analysis highlighting the development over time of scientific contributions in this field. Our findings show that the academic discourse has gained momentum but is still pre-mature. Additionally, we find that a technical perspective on the topic prevails in literature.

In the digital era, platform work has become prevalent, with millions of individuals striving to become influencers on popular social media platforms, like YouTube and Instagram. However, despite its benefits, platform work ex-poses content creators to the pervasive influence of algorithms that exert control over their activities. With content creation rapidly developing into a booming industry, there is a pressing need to better understand content creators’ percep-tions of platform control, including its dimensionality and implications for crea-tors’ well-being, performance, and creativity. Developed as part of a larger re-search project, we present the initial steps of the scale development process for platform control in this research-in-progress paper. This initial step contributes to a better understanding of the evolving landscape of digital work and its broader implications for individuals and society.