Don't Go Breaking My Trust: A Taxonomy of Interaction Failures for Transparent Robots

Beatriz Nogueira, João Ferreira, Hugo Simão, Filipa Rocha, Isabel Neto, João Guerreiro, Tiago Guerreiro

Keywords: HRI Taxonomy, Trust, Transparency, Failures

TL;DR: This paper presents a multidimensional taxonomy of human-robot interaction failures that classifies breakdowns across perception, cognition, and execution to equip robots with targeted, context-aware explanations and preserve user trust.

Abstract: As robots become integrated into everyday environments, interaction failures are inevitable. However, current robotics design often treats failures as technical bugs rather than interactive mistakes. When these errors are presented in technical terms, users might not be able to understand them, reducing trust and the ease of the interaction. To address this, we present a preliminary, multidimensional taxonomy of HRI failures. Based on team brainstorming and literature, our framework classifies failures across three core system capabilities: Perception, Cognition, and Execution. Crucially, it attributes these failures not solely to the robot, but also to human and environmental factors, while assessing their impact on user trust and safety. By pinpointing the exact nature and origin of an error, this taxonomy provides the vocabulary needed to equip robots with targeted, context-aware explanations, advancing the design of transparent and understandable human-robot interaction.

Submission Number: 5