Master Projects (open)

A Master Project is a group project with a concrete topic of limited scope that is attached to a research group, such as the IVDA group. The project is conducted by 2-4 students. It includes challenging and intensive work and is worth 15 ECTS. Students must choose a topic that suits their major track. We currently have several topics available (see list in the menu).

To apply, please reach out to the relevant supervisors (as listed in the project pages) with your: CV, transcripts, and a short statement about why you are suited to this project in particular, and what about it appeals to you.

A Master Project Proposal for UZH Master’s Students in Informatics

The Problem

Meetings are where collaborative sensemaking happens or fails. Teams align (or misalign) on problems, decisions, and next steps through conversation. Yet much of what gets said disappears: partial attention, note-taking gaps, differing interpretations, and topic drift mean that after a meeting, participants often leave with inconsistent mental models of what was decided, why, and what it means. Existing AI meeting tools help, but they are fundamentally passive and retrospective. Microsoft Teams Facilitator, for example, generates live notes, tracks agenda items, summarizes decisions, and flags open questions — all useful, but structured as a text-based log that participants read about, not a model they think with. The representation stays one step removed from the sensemaking process itself. What if AI support were visual, live, and interactive — not a transcript running alongside the meeting, but a shared conceptual space that evolves with the discussion and gives all participants a common ground to reason from, in real time?

The Idea: A Shared Live Mental Model

We propose building a prototype AI-powered sensemaking interface that listens to a meeting in real time, continuously constructs a structured visual model of the discussion state, and presents that model to all participants as a shared, interactive artifact on a shared or individual display.
The system would:
–    Listen to the meeting audio (voice-to-text with speaker attribution)
–    Model the evolving discussion, e.g., key topics, open questions, decisions, perspectives, gaps, and contradictions (essential features to be determined in the project). This model should be continuously maintained and updated with new data
–    Visualize this as an interactive shared representation participants can view during the meeting
–    Enable direct interaction: participants can annotate, highlight, redirect attention, or flag items on the model while the discussion is ongoing
This goes beyond transcript-style tools in two important ways. First, the representation is structural and visual, designed to surface latent relationships and reduce cognitive load rather than add to it. Second, it is interactive and shared — a common artifact the group uses to steer conversation and verify alignment, not just review afterward.

Empirical Research Questions (work in progress)

The prototype is a probe that opens a cluster of potential high-value empirical questions, e.g. :
–    Sensemaking and alignment: Does a live shared visual model help participants develop more aligned understanding of the discussion? Or does it create an illusion of alignment that masks real divergence? Will participants share more unique/controversial/diverse contributions or less?
–    Focus vs. Distraction : Will the display help groups stay on topic, e.g., through making drift visible as it happens, or will it pull attention away from the human conversation and toward managing the model? (see also Riedl et al. 2026 for attention)
–    Result characteristics : Will the access to interactive model lead to more detailed, coherent, creative, feasible… results of the meeting ? (depending on the meeting goal and model features)
–    Intervention design: What proactive signals should the system generate? E.g., surfacing an uncovered agenda item, flagging a contradiction, drawing attention to a neglected perspective — when are these helpful nudges versus unwanted interruptions?
–    Interface and interaction design: How do you represent a dynamic, multi-threaded discussion visually without overwhelming participants (cognitive load)? What interaction patterns work in co-located versus virtual settings? What is a good shared (visual) reference? Is this dynamically created or pre-defined based on the meeting type (of course, the former would be better).