Department of Mathematics and Computer Science

AI systems increasingly adapt their behavior to individual users. However, the terminology used to describe such individualized systems remains inconsistent. Terms such as personalization, customization, adaptation, adaptivity, and adaptability are often used to describe related but conceptually different forms of individualization. Classical HCI research, for example, distinguishes between system-controlled adaptive interfaces and user-controlled adaptable interfaces, while mixed-initiative systems distribute control between users and systems.

With the emergence of Large Language Models (LLMs), these distinctions become increasingly relevant: AI behavior can be configured explicitly by users, inferred automatically by the system, negotiated interactively, or adjusted continuously during interaction. At the same time, recent work uses different terminology to describe these mechanisms.

This thesis aims to systematically map how different research communities conceptualize and operationalize individualized AI. The goal is to develop a structured vocabulary that distinguishes different forms of individualization according to dimensions such as who initiates an adaptation, who has control over it, what is adapted, and when adaptation occurs.

Procedure

The thesis follows a scoping review methodology. In close collaboration with the advisors, an initial research question and search strategy are developed to capture relevant terminology across HCI, Human-AI Interaction, and related research on LLM-based and adaptive systems. Following established scoping review guidelines, relevant literature is systematically identified, screened, and analyzed.

For the included publications, terminology and underlying concepts are extracted and coded. Particular attention is paid to how papers define concepts such as personalization, customization, and adaptation, and to dimensions such as distribution of control, user involvement, timing, and adaptation target. The resulting conceptual landscape is synthesized into a taxonomy or terminology framework for individualized AI and LLM-based systems.

The thesis is conducted in close collaboration with the research group, particularly during scope definition, development of the coding scheme, and synthesis of the resulting terminology. 

References

• Horvitz, E. (1999). Principles of Mixed-Initiative User Interfaces. Proceedings of CHI '99.
• Findlater, L., & McGrenere, J. (2004). A Comparison of Static, Adaptive, and Adaptable Menus. Proceedings of CHI '04.
• Houde, S., et al. (2025). Controlling AI Agent Participation in Group Conversations: A Human-Centered Approach. Proceedings of IUI '25.
• Vossen, W., Szymanski, M., & Verbert, K. (2024). The Effect of Personalizing a Psychotherapy Conversational Agent on Therapeutic Bond and Usage Intentions. Proceedings of IUI '24.
• Tricco, A. C., et al. (2018). PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Annals of internal medicine, 169(7), 467–473. https://doi.org/10.7326/M18-0850