The Bystander Affect Detection (BAD) Dataset for Failure Detection in HRI

TL;DR

This paper introduces the BAD dataset of bystander reactions to robot and human errors, and demonstrates a deep learning model achieving over 90% precision in failure detection, advancing robot error awareness.

Contribution

The paper presents a novel dataset of bystander reactions and evaluates deep learning methods for failure detection in human-robot interaction scenarios.

Findings

Deep learning model achieved over 90% precision in failure detection.

Different data labeling methods significantly impact model performance.

The BAD dataset supports future research in error detection and response modeling.

Abstract

For a robot to repair its own error, it must first know it has made a mistake. One way that people detect errors is from the implicit reactions from bystanders -- their confusion, smirks, or giggles clue us in that something unexpected occurred. To enable robots to detect and act on bystander responses to task failures, we developed a novel method to elicit bystander responses to human and robot errors. Using 46 different stimulus videos featuring a variety of human and machine task failures, we collected a total of 2452 webcam videos of human reactions from 54 participants. To test the viability of the collected data, we used the bystander reaction dataset as input to a deep-learning model, BADNet, to predict failure occurrence. We tested different data labeling methods and learned how they affect model performance, achieving precisions above 90%. We discuss strategies to model bystander reactions and predict failure and how this approach can be used in real-world robotic deployments to detect errors and improve robot performance. As part of this work, we also contribute with the "Bystander Affect Detection" (BAD) dataset of bystander reactions, supporting the development of better prediction models.