Can Robots Explain Their Mistakes? (LLMs in Robotics)

Robots fail—and how they explain those failures can make or break trust.
Holly Yanco shares research from her lab on robot failure explanations: starting with behavior trees + assumption checkers, then testing LLM-generated explanations. Early results? People didn’t understand the LLM outputs… until the team prompted for novices and grounded the explanation with environmental context (like an image).

Big takeaway: LLMs can help, but correctness and hallucinations are a real safety concern.

#Robotics #AI #LLM #HumanRobotInteraction #HRI #TrustInAI