Learning Personal Topic Relevance in Remote Meetings with GenAI

Problem Description

Remote meetings have become a central part of everyday knowledge work, yet participants often face competing demands and multitask during discussions. As a result, they frequently miss content that may be important for them and struggle to keep track of topics that matter for their role or responsibilities. At the same time, recent research and industry tools increasingly explore adaptive meeting support, such as personalized alerts or tailored summaries. However, a key challenge remains unresolved: How can a system reliably infer which meeting topics are relevant for a specific person? While prior work has proposed manual preference selection, there is little understanding of whether behavioral interaction patterns can be used to learn personal relevance profiles over time. This thesis addresses this gap by exploring whether multimodal interaction data collected across multiple meetings can serve as indicators of personal topic relevance.

Goal of Thesis

The primary goal of this thesis is to identify, design, and evaluate a system that infers personal topic relevance in remote meetings using multimodal interaction patterns collected across multiple meetings. Following a Design Science Research (DSR) approach, the thesis will:

  1. Review theoretical foundations from remote meeting research, multitasking, attention, engagement, and agentic information systems.
  2. Identify which interaction data sources (e.g., speaking activity, window focus, eye tracking) can indicate personal topic relevance in remote meetings and derive design requirements based on theory and user needs.
  3. Implement a prototype based on the design requirements that collects selected interaction data across multiple meetings, segments meeting content into topics using AI, and infers personal topic relevance.
  4. Evaluate the accuracy and usefulness of the relevance predictions through a field study with users.
Requirements
  • Strong programming skills and experience with modern software development (e.g., Electron, JavaScript/TypeScript).
  • Familiarity with Generative AI and natural language processing (e.g., topic segmentation, summarization).
  • Understanding of the Design Science Research methodology and ability to conduct user evaluations.
  • Strong time management and communication skills and proficiency in English.