@article{65105, author = {{zur Heiden, Philipp and Halimeh, Haya and Hansmeier, Philipp and Vorbohle, Christian and Althaus, Maike and Beverungen, Daniel and Kundisch, Dennis and Müller, Oliver}}, journal = {{Communications of the Association for Information Systems}}, title = {{{Data Spaces for Heterogeneous Data Ecosystems – Findings from a Design Study in the Cultural Sector}}}, year = {{2026}}, } @article{65810, author = {{Zapata Gonzalez, David Ricardo and Meyer, Marcel and Zalipski, Kevin and Müller, Oliver}}, issn = {{0306-2619}}, journal = {{Applied Energy}}, publisher = {{Elsevier BV}}, title = {{{Temporal causal feature selection for robust machine learning modelling of data center operations}}}, doi = {{10.1016/j.apenergy.2026.127984}}, volume = {{417}}, year = {{2026}}, } @inproceedings{66903, author = {{Meyer, Marcel and Kaltenpoth, Sascha Benjamin and Albers, Henrik and Zalipski, Kevin and Müller, Oliver}}, booktitle = {{Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2}}, publisher = {{ACM}}, title = {{{TS-Arena -- A Live Forecast Pre-Registration Platform}}}, doi = {{10.1145/3770855.3817515}}, year = {{2026}}, } @inbook{58724, author = {{Brennig, Katharina and Kaltenpoth, Sascha Benjamin and Müller, Oliver}}, booktitle = {{Lecture Notes in Business Information Processing}}, isbn = {{9783031786655}}, issn = {{1865-1348}}, publisher = {{Springer Nature Switzerland}}, title = {{{Straight Outta Logs: Can Large Language Models Overcome Preprocessing in Next Event Prediction?}}}, doi = {{10.1007/978-3-031-78666-2_15}}, year = {{2025}}, } @inproceedings{63524, abstract = {{Recommendation systems are essential for delivering personalized content across e-commerce and streaming services. However, traditional methods often fail in cold-start scenarios where new items lack prior interactions. Recent advances in large language models (LLMs) offer a promising alternative. In this paper, we adopt the retrieve-and-recommend framework and propose to fine-tune the LLM jointly on warm-and cold-start next-item recommendation tasks, thus, mitigating the need for separate models for both item types. We computationally compare zero-shot prompting, in-context learning, and fine-tuning using the same LLM backbone, and benchmark them against strong PLM-based baselines. Our findings provide practical insights into the trade-offs between accuracy and computational cost of these methods for next-item recommendation. To enhance reproducibility, we release the source code under https://github. com/HayaHalimeh/LLMs-For-Next-Item-Recommendation.git.}}, author = {{Halimeh, Haya and Freese, Florian and Müller, Oliver}}, booktitle = {{International Conference on Information Systems Development}}, issn = {{2938-5202}}, publisher = {{University of Gdansk, Department of Business Informatics & University of Belgrade, Faculty of Organizational Sciences}}, title = {{{LLMs For Warm and Cold Next-Item Recommendation: A Comparative Study across Zero-Shot Prompting, In-Context Learning and Fine-Tuning}}}, doi = {{10.62036/isd.2025.68}}, year = {{2025}}, } @inproceedings{63523, abstract = {{Data spaces have become a strategic pillar of Europe's digital agenda, enabling sovereign, legally compliant data sharing within decentralized ecosystems. As data space initiatives evolve, personalized recommendations are increasingly recognized as key use cases. However, traditional recommendation approaches typically rely on centralized aggregation of user behavior data-directly conflicting with the core ethos of data spaces: sovereignty, privacy, and trust. Federated recommendation systems offer a promising alternative by training models locally and exchanging only intermediate parameters to build a global model. Despite this potential, the integration of federated recommendation techniques and data space architectures remains largely underexplored in research and practice. This paper addresses this gap by designing and evaluating a prototype of a federated recommendation system specifically tailored for data spaces and compliant with their underlying infrastructure. Our findings highlight the viability of developing privacy-preserving, collaborative recommendation systems within data spaces, and contribute to the broader adoption of AI across these emerging ecosystems.}}, author = {{Halimeh, Haya and zur Heiden, Philipp}}, booktitle = {{2025 27th International Conference on Business Informatics (CBI)}}, publisher = {{IEEE}}, title = {{{Preserving Sovereignty and Privacy for Personalization: Designing a Federated Recommendation System for Data Spaces}}}, doi = {{10.1109/cbi68102.2025.00019}}, year = {{2025}}, } @inproceedings{63525, abstract = {{Recommender systems (RS) can support sustainable development by steering users toward more sustainable choices. Sustainability-aware explanations represent one avenue for contributing to this goal by foregrounding the environmental and social aspects of the recommended products or services. This paper advances the line of research on sustainability-aware explanations by integrating nudging mechanisms into their design and by evaluating their effectiveness through a randomized between-subjects online vignette experiment across two item domains (). Our findings offer actionable design guidelines for building RS that foster sustainability-aware decision making and enrich the empirical foundation for impact-oriented research on explanation in RS. }}, author = {{Halimeh, Haya and Müller, Oliver}}, location = {{Prague, Czech Republic}}, title = {{{Towards Greener Choices: Decision Information Nudging for Sustainability-Aware Recommender Explanations}}}, doi = {{10.1007/978-3-032-13342-7}}, year = {{2025}}, } @techreport{63026, author = {{Althaus, Maike and Beverungen, Daniel and Flath, Beate and Halimeh, Haya and Hansmeier, Philipp and zur Heiden, Philipp and Kundisch, Dennis and Müller, Michelle and Müller, Oliver and Oberthür, Simon and Vorbohle, Christian and Momen Pour Tafreshi, Maryam and Mauß, Sebastian and Mücke, Alina and Müller, Jörg and Peter, Malte and Schmitt-Chandon, Ariane and Sellerberg, Kerstin and Steinhäuser, Moritz}}, title = {{{Positionspapier Use Case 1: Vernetzte Kulturplattformen}}}, year = {{2025}}, } @inproceedings{60680, abstract = {{Classical machine learning techniques often struggle with overfitting and unreliable predictions when exposed to novel conditions. Introducing causality into the modelling process offers a promising way to mitigate these challenges by enhancing predictive robustness. However, constructing an initial causal graph manually using domain knowledge is time-consuming, particularly in complex time series with numerous variables. To address this, causal discovery algorithms can provide a preliminary causal structure that domain experts can refine. This study investigates causal feature selection with domain knowledge using a data center system as an example. We use simulated time-series data to compare different causal feature selection with traditional machine-learning feature selection methods. Our results show that predictions based on causal features are more robust compared to those derived from traditional methods. These findings underscore the potential of combining causal discovery algorithms with human expertise to improve machine learning applications.}}, author = {{Zapata Gonzalez, David Ricardo and Meyer, Marcel and Müller, Oliver}}, keywords = {{Causal Machine Learning, Causality in Time Series, Causal Discovery, Human-Machine Collaboration}}, location = {{Amman, Jordan}}, title = {{{Bridging the gap between data-driven and theory-driven modelling – leveraging causal machine learning for integrative modelling of dynamical systems}}}, year = {{2025}}, } @inproceedings{60958, abstract = {{Large Language Models (LLMs) excel in understanding, generating, and processing human language, with growing adoption in process mining. Process mining relies on event logs that capture explicit process knowledge; however, knowledge-intensive processes (KIPs) in domains such as healthcare and product development depend on tacit knowledge, which is often absent from event logs. To bridge this gap, this study proposes a LLM-based framework for mobilizing tacit process knowledge and enriching event logs. A proof-of-concept is demonstrated using a KIP-specific LLM-driven conversational agent built on GPT-4o. The results indicate that LLMs can capture tacit process knowledge through targeted queries and systematically integrate it into event logs. This study presents a novel approach combining LLMs, knowledge management, and process mining, advancing the understanding and management of KIPs by enhancing knowledge accessibility and documentation.}}, author = {{Brennig, Katharina}}, booktitle = {{AMCIS 2025 Proceedings. 11.}}, keywords = {{Process Mining, Large Language Model, Knowledge Management, Knowledge-Intensive Process, Tacit Knowledge}}, location = {{Montréal}}, title = {{{Revealing the Unspoken: Using LLMs to Mobilize and Enrich Tacit Knowledge in Event Logs of Knowledge-Intensive Processes}}}, year = {{2025}}, } @inproceedings{63400, abstract = {{Data centers (DCs) form the backbone of our growing digital economy, but their rising energy demands pose challenges to our environment. At the same time, reusing waste heat from DCs also represents an opportunity, for example, for more sustainable heating of residential buildings. Modeling and optimizing these coupled and dynamic systems of heat generation and reuse is complex. On the one hand, physical simulations can be used to model these systems, but they are time-consuming to develop and run. Machine learning (ML), on the other hand, allows efficient data-driven modeling, but conventional correlation-based approaches struggle with the prediction of interventions and out-of-distribution generalization. Recent advances in causal ML, which combine principles from causal inference with flexible ML methods, are a promising approach for more robust predictions. Due to their focus on modeling interventions and cause-and-effect relationships, it is difficult to evaluate causal ML approaches rigorously. To address this challenge, we built a testbed of a miniature DC with an integrated waste heat network, equipped with sensors and actuators. This testbed allows conducting controlled experiments and automatic collection of realistic data, which can then be used to benchmark conventional and causal ML methods. Our experimental results highlight the strengths and weaknesses of each modeling approach, providing valuable insights on how to appropriately apply different types of machine learning to optimize data center operations and enhance their sustainability.}}, author = {{Zapata Gonzalez, David Ricardo and Meyer, Marcel and Müller, Oliver}}, booktitle = {{ACM SIGEnergy Energy Informatics Review}}, issn = {{2770-5331}}, number = {{2}}, pages = {{4--10}}, publisher = {{Association for Computing Machinery (ACM)}}, title = {{{Causal Machine Learning Approaches for Modelling Data Center Heat Recovery: A Physical Testbed Study}}}, doi = {{10.1145/3757892.3757893}}, volume = {{5}}, year = {{2025}}, } @article{64870, author = {{Meyer, Marcel and Zapata Gonzalez, David Ricardo and Kaltenpoth, Sascha Benjamin and Müller, Oliver}}, issn = {{2169-3536}}, journal = {{IEEE Access}}, pages = {{218141--218153}}, publisher = {{Institute of Electrical and Electronics Engineers (IEEE)}}, title = {{{Benchmarking Time Series Foundation Models for Short-Term Household Electricity Load Forecasting}}}, doi = {{10.1109/access.2025.3648056}}, volume = {{13}}, year = {{2025}}, } @article{53130, author = {{Stumpe, Miriam and Dieter, Peter and Schryen, Guido and Müller, Oliver and Beverungen, Daniel}}, journal = {{Transportation Research Part A: Policy and Practice}}, title = {{{Designing taxi ridesharing systems with shared pick-up and drop-off locations: Insights from a computational study}}}, year = {{2024}}, } @inproceedings{55096, author = {{Bösch, Kevin and Müller, Oliver and Weinmann, Markus}}, booktitle = {{Proceedings of the Symposium on Statistical Challenges in Electronic Commerce Research}}, title = {{{Not your Average Digital Nudge: Heterogeneous Effects of Personalized Nudges with CausalML}}}, year = {{2024}}, } @inproceedings{56945, abstract = {{Adopting Large language models (LLMs) in organizations potentially revolutionizes our lives and work. However, they can generate off-topic, discriminating, or harmful content. This AI alignment problem often stems from misspecifications during the LLM adoption, unnoticed by the principal due to the LLM’s black-box nature. While various research disciplines investigated AI alignment, they neither address the information asymmetries between organizational adopters and black-box LLM agents nor consider organizational AI adoption processes. Therefore, we propose LLM ATLAS (LLM Agency Theory-Led Alignment Strategy) a conceptual framework grounded in agency (contract) theory, to mitigate alignment problems during organizational LLM adoption. We conduct a conceptual literature analysis using the organizational LLM adoption phases and the agency theory as concepts. Our approach results in (1) providing an extended literature analysis process specific to AI alignment methods during organizational LLM adoption and (2) providing a first LLM alignment problem-solutionspace.}}, author = {{Kaltenpoth, Sascha Benjamin and Müller, Oliver}}, booktitle = {{Wirtschaftsinformatik 2024 Proceedings}}, title = {{{Getting in Contract with Large Language Models - An Agency Theory Perspective On Large Language Model Alignment}}}, year = {{2024}}, } @inproceedings{37312, abstract = {{Optimal decision making requires appropriate evaluation of advice. Recent literature reports that algorithm aversion reduces the effectiveness of predictive algorithms. However, it remains unclear how people recover from bad advice given by an otherwise good advisor. Previous work has focused on algorithm aversion at a single time point. We extend this work by examining successive decisions in a time series forecasting task using an online between-subjects experiment (N = 87). Our empirical results do not confirm algorithm aversion immediately after bad advice. The estimated effect suggests an increasing algorithm appreciation over time. Our work extends the current knowledge on algorithm aversion with insights into how weight on advice is adjusted over consecutive tasks. Since most forecasting tasks are not one-off decisions, this also has implications for practitioners.}}, author = {{Leffrang, Dirk and Bösch, Kevin and Müller, Oliver}}, booktitle = {{Hawaii International Conference on System Sciences}}, keywords = {{Algorithm aversion, Time series, Decision making, Advice taking, Forecasting}}, title = {{{Do People Recover from Algorithm Aversion? An Experimental Study of Algorithm Aversion over Time}}}, year = {{2023}}, } @inproceedings{50121, abstract = {{Many researchers and practitioners see artificial intelligence as a game changer compared to classical statistical models. However, some software providers engage in “AI washing”, relabeling solutions that use simple statistical models as AI systems. By contrast, research on algorithm aversion unsystematically varied the labels for advisors and treated labels such as "artificial intelligence" and "statistical model" synonymously. This study investigates the effect of individual labels on users' actual advice utilization behavior. Through two incentivized online within-subjects experiments on regression tasks, we find that labeling human advisors with labels that suggest higher expertise leads to an increase in advice-taking, even though the content of the advice remains the same. In contrast, our results do not suggest such an expert effect for advice-taking from algorithms, despite differences in self-reported perception. These findings challenge the effectiveness of framing intelligent systems as AI-based systems and have important implications for both research and practice.}}, author = {{Leffrang, Dirk}}, booktitle = {{International Conference on Information Systems}}, keywords = {{Artificial Intelligence, Algorithm Appreciation, Framing, Advice-taking, Expertise}}, location = {{Hyderabad, India}}, number = {{10}}, title = {{{AI Washing: The Framing Effect of Labels on Algorithmic Advice Utilization}}}, year = {{2023}}, } @inproceedings{50118, abstract = {{Despite the widespread use of machine learning algorithms, their effectiveness is limited by a phenomenon known as algorithm aversion. Recent research concluded that unobserved variables can cause algorithm aversion. However, the impact of an unobserved variable on algorithm aversion remains unclear. Previous studies focused on situations where humans had more variables available than algorithms. We extend this research by conducting an online experiment with 94 participants, systematically varying the number of observable variables to the advisor and the advisor type. Surprisingly, our results did not confirm that an unobserved variable had a negative effect on advice-taking. Instead, we found a positive impact in an algorithm appreciation scenario. This study provides new insights into the paradoxical behavior in which people weigh advice more despite having fewer variables, as they correct for the advisor's errors. Practitioners should consider this behavior when designing algorithms and account for user correction behavior.}}, author = {{Leffrang, Dirk}}, booktitle = {{Wirtschaftsinformatik Conference}}, keywords = {{Algorithm aversion, Data, Decision-making, Advice-taking, Human-Computer Interaction}}, location = {{Paderborn}}, number = {{19}}, title = {{{The Broken Leg of Algorithm Appreciation: An Experimental Study on the Effect of Unobserved Variables on Advice Utilization}}}, year = {{2023}}, } @inproceedings{50431, abstract = {{Recommender systems now span the entire customer journey. Amid the multitude of diversified experi- ences, immersing in cultural events has become a key aspect of tourism. Cultural events, however, suffer from fleeting lifecycles, evade exact replication, and invariably lie in the future. In addition, their low standardization makes harnessing historical data regarding event content or past patron evaluations intricate. The distinctive traits of events thereby compound the challenge of the cold-start dilemma in event recommenders. Content-based recommendations stand as a viable avenue to alleviate this issue, functioning even in scenarios where item-user information is scarce. Still, the effectiveness of content- based recommendations often hinges on the quality of the data representation they build upon. In this study, we explore an array of cutting-edge uni- and multimodal vision and language foundation models (VL-FMs) for this purpose. Next, we derive content-based recommendations through a straightforward clustering approach that groups akin events together, and evaluate the efficacy of the models through a series of online user experiments across three dimensions: similarity-based evaluation, comparison-based evaluation, and clustering assignment evaluation. Our experiments generated four major findings. First, we found that all VL-FMs consistently outperformed a naive baseline of recommending randomly drawn events. Second, unimodal text-based embeddings were surprisingly on par or in some cases even superior to multimodal embeddings. Third, multimodal embeddings yielded arguably more fine-grained and diverse clusters in comparison to their unimodal counterparts. Finally, we could confirm that cross event interest is indeed reliant on the perceived similarity of events, resonating with the notion of similarity in content-based recommendations. All in all, we believe that leveraging the potential of contemporary FMs for content-based event recommendations would help address the cold-start problem and propel this field of research forward in new and exciting ways.}}, author = {{Halimeh, Haya and Freese, Florian and Müller, Oliver}}, booktitle = {{Workshop on Recommenders in Tourism, co-located with the 17th ACM Conference on Recommender Systems}}, title = {{{Event Recommendations through the Lens of Vision and Language Foundation Models}}}, year = {{2023}}, } @inproceedings{45270, abstract = {{Clinical depression is a serious mental disorder that poses challenges for both personal and public health. Millions of people struggle with depression each year, but for many, the disorder goes undiagnosed or untreated. Over the last decade, early depression detection on social media emerged as an interdisciplinary research field. However, there is still a gap in detecting hesitant, depression-susceptible individuals with minimal direct depressive signals at an early stage. We, therefore, take up this open point and leverage posts from Reddit to fill the addressed gap. Our results demonstrate the potential of contemporary Transformer architectures in yielding promising predictive capabilities for mental health research. Furthermore, we investigate the model’s interpretability using a surrogate and a topic modeling approach. Based on our findings, we consider this work as a further step towards developing a better understanding of mental eHealth and hope that our results can support the development of future technologies.}}, author = {{Halimeh, Haya and Caron, Matthew and Müller, Oliver}}, booktitle = {{Hawaii International Conference on System Sciences}}, keywords = {{Social Media and Healthcare Technology, early depression detection, liwc, mental health, transfer learning, transformer architectures}}, title = {{{Early Depression Detection with Transformer Models: Analyzing the Relationship between Linguistic and Psychology-Based Features}}}, year = {{2023}}, }