Research · Study 02
Trust Dynamics Across Multiple Robot Failures
How repeated failures and failure awareness shape user trust over time
Robot failures often happen more than once during collaboration, and their effects on user trust can accumulate. I examine how repeated failures change trust, perceived robot intelligence, and how much it helps for a robot to show awareness of its own failures.
Key Takeaways
- User trust is influenced by both the current failure and the failures that came before it.
- Different failure sequences with similar severity can produce different trust and intelligence judgments.
- Failure awareness helps after serious failures, but it can reduce trust when the failure is minor or barely noticeable.
Demo
Overview
This study explores the dynamics of trust in human-robot collaboration in the context of repeated robot failures. It focuses on how the consistency of failure types and the robot's awareness of its own errors affect user perceptions of trust and intelligence.
Method and System
Using a collaborative Tangram puzzle task, I ran a mixed-design experiment with 54 participants. The robot intentionally failed under different conditions, manipulating both the failure sequence type (homogeneous vs. heterogeneous) and the robot's level of failure awareness (none, partial, full) across the failure types Freezing, Planning, and Grasping.
What I Found
User trust and perceived intelligence are affected by both the current and the previous failure, with homogeneous sequences generally causing a smaller trust decline than heterogeneous ones. Robots that demonstrated awareness — full or partial — were rated significantly higher in trust and perceived intelligence than those showing no awareness, particularly for grasping and planning failures.
Why It Matters
These insights offer design implications for building more resilient and trustworthy human-robot interactions amid inevitable errors. Knowing when failure awareness helps — and when it backfires — lets us design adaptive trust-repair strategies rather than one-size-fits-all apologies.