Open Topics

An overview of possible thesis topic(s) currently offered by our lab.

Topic 1

LLMs as Mediators: Consensus Finding and Conflict Highlighting in Group Recommender Systems

Classical group recommenders aggregate preferences with static strategies (average, least misery, most pleasure), which conceal disagreement rather than resolve it. LLMs open a new design space: they can elicit preferences conversationally, surface conflicts explicitly, propose compromise items with natural-language justifications, and adapt aggregation dynamically. This thesis addresses mediation and preference-balancing aspects of using LLMs in group recommender systems.

Possible Methodology:

  1. Literature review on group recommendation aggregation strategies, computational negotiation/mediation, and explainable recommendation. Creation of research questions.
  2. Design and implementation of a hybrid system e.g. a conventional recommender backbone (e.g., collaborative filtering over a shopping/movies dataset) combined with an LLM mediation layer that manages group dialogue, polling, and conflict visualization.
  3. Design and execution of a suitable evaluation strategy for the proposed system. Students are expected to identify appropriate baselines, evaluation methods, and metrics, and to justify how the quality of the aggregation and the effect of different mediation behaviors on the group can be measured in a sound way.

Possible Contributions:

A working dynamic LLM-based group recommender prototype and empirical evidence on which mediation strategies actually improve group harmony rather than merely masking disagreement.

Possible Relevant Literature:

  • Wagner, Ahmadou & Kolb, Thomas & Banerjee, Ashmi & Nazary, Fatemeh & Neidhardt, Julia & Deldjoo, Yashar. (2025). Conversational Recommender Systems Using Generative Models (Gen-CRS): A Literature Review. https://doi.org/10.1145/3828551.
  • Dara, S., Chowdary, C.R. & Kumar, C. (2020). A survey on group recommender systems. Journal of Intelligent Information Systems 54, 271–295. https://doi.org/10.1007/s10844-018-0542-3
  • Lubos, Sebastian, et al. (2025). Towards LLM-Enhanced Group Recommender Systems. arXiv preprint arXiv:2507.19283.

Topic 2

When to Ask the Machine: Optimal Timing of AI Recommender Interventions in Group Decision-Making

Recommender systems are typically evaluated with respect to what they recommend, whereas when a recommendation is introduced into an ongoing group decision process remains largely unexplored. A suggestion offered too early can anchor the group and suppress genuine preference elicitation; offered too late, it may be ignored or perceived as disruptive of an emerging consensus. This thesis addresses the identification of key decision points and the optimal-timing analysis aspects of introducing AI recommenders into group decision-making, studying how the stage of intervention affects decision quality, satisfaction, and group harmony.

Possible Methodology:

  1. Literature review on group decision processes and consensus building, decision-stage models, and anchoring/conformity effects in human–AI interaction. Operationalization of decision stages (preference formation, negotiation, convergence) and creation of research questions.
  2. Design and implementation of an experimental group-shopping environment (or extension of an existing CDL prototype) in which an LLM-based recommender can be injected at configurable decision points, with polling and stage-detection instrumentation.
  3. Design and execution of a suitable evaluation strategy to assess how the timing of the intervention affects the group and its decision. Students are expected to determine appropriate study designs, conditions, and metrics, and to justify how the effect of different intervention timings on decision quality and group outcomes can be measured in a sound way.

Possible Contributions:

An empirically grounded model of decision-stage-dependent recommender effectiveness, together with concrete design guidelines on when to introduce AI into group decision processes — directly feeding the best-practices and timings report foreseen as the second WP3 deliverable.

Possible Relevant Literature:

  • Adomavicius, G., Bockstedt, J.C., Curley, S.P. & Zhang, J. (2013). Do Recommender Systems Manipulate Consumer Preferences? A Study of Anchoring Effects. Information Systems Research 24(4), 956–975. https://doi.org/10.1287/isre.2013.0497
  • Masthoff, J. (2015). Group Recommender Systems: Aggregation, Satisfaction and Group Attributes. In: Recommender Systems Handbook, Springer, 743–776. https://doi.org/10.1007/978-1-4899-7637-6_22
  • Delic, A., Masthoff, J., Neidhardt, J., et al. (2024). An overview of consensus models for group decision-making and group recommender systems. User Modeling and User-Adapted Interaction 34, 489–547. https://doi.org/10.1007/s11257-023-09380-z

How to Apply: 

Both topics are offered within the Christian Doppler Laboratory for Recommender Systems at TU Wien and are scoped for approximately 30 ECTS. If you are interested in any of these topics, please email us at recsys-lab[ät]ec.tuwien.ac.at  and include the following documents:

  • a short motivational letter (max. 1 page) indicating the preferred topic
  • a curriculum vitae (CV)
  • a transcript of grades