Real Time Detection of Blind Spots in Group Discussions - User Evaluation of an AI Based Insight Tool

This thesis is embedded in an applied research collaboration between UZH, ZHAW (Prof. de Spindler) and an industry partner. An established facilitation method used in organisational training aims to help teams gain clarity through so called “playful boosters”. A new prototype will be developed, which extends this method with AI-based analysis of ongoing conversations. It highlights missing viewpoints, underrepresented themes, and subtle shifts in group alignment. This thesis focuses on testing how well these cues work for users who experience the prototype in action.

You will run small scale evaluations with student or organisational teams and assess how they interpret the system’s feedback. The results will help refine how the prototype surfaces blind spots and how these signals should be presented.

Possible tasks include

• Designing and conducting user tests with real team interactions
• Collecting feedback on clarity, trust, and usefulness
• Observing when users act on detected blind spots
• Comparing different prompt or display variants
• Reporting strengths, weaknesses, and opportunities for refinement

Information systems, human computer interaction, applied AI, psychology of interaction, or related fields.

• Interest in human AI collaboration and conversational interfaces
• Basic skills in qualitative study design
• Motivation to work with early-stage prototypes
• German skills are helpful

For more information please contact Prof. Dr. Gerhard Schwabe