Reactive robots beat proactive ones in group escape room study

A team of researchers led by Thomas Vitry at the University of Hamburg investigated how a robot's initiative affects multi-party human-robot collaboration. They designed an escape room where pairs of participants worked with a humanoid robot under two interaction models: a reactive model, where the robot only responded when directly addressed, and a proactive model, where the robot continuously listened, contributed autonomously, and periodically re-initiated interaction. The study used puzzle-solving performance, interaction frequency, and participant ratings on the Godspeed and RoSAS scales to evaluate both approaches.

The results revealed a clear trade-off. The proactive model substantially increased interaction frequency, but the reactive model achieved a descriptively higher overall success rate—92.86% versus 71.42%. The strongest differences emerged when accounting for prior experience and personality: participants with LLM experience solved early puzzles faster in the reactive condition, while those with prior robot experience or introverted traits showed modified evaluations of proactive vs. reactive interaction. These findings, published at the RO-MAN 2026 conference, demonstrate that robot initiative effects are shaped by users' prior experience, personality traits, and group needs.