Publication
Causal explanations for robot decisions and beliefs
Conference Article
Conference
ACM/IEEE International Conference on Human-Robot Interaction (HRI)
Edition
2026
Pages
1342-1344
Doc link
https://doi.org/10.1145/3776734.3794634
File
Abstract
Explainability is an important tool for human-robot interaction (HRI). By explaining its decisions and beliefs, a robot can promote understandability and thereby foster desiderata such as trust, acceptance and usability. However, HRI domains pose challenges to automatic explanation generation. In such domains, a robot must consider the causal reasons for behaviour embedded in temporal sequences of decisions, all while factoring in noise and uncertainty inherent to these kinds of domains. Additionally, as explainability itself constitutes a human-robot interaction, it is important for robots to be able to properly interpret user questions and effectively communicate explanations in order to improve understanding. In our work, we address these challenges from a causal perspective, developing methods that use causal models in order to automatically generate causal, counterfactual explanations in HRI domains. We also produce some insights into embedding such a system in a human-robot interaction in order to maximise understandability.
Categories
knowledge engineering, social aspects of automation.
Author keywords
Explainability, Human-Robot Interaction, Causal Models, Counterfactual Explanations
Scientific reference
T. Love, A. Andriella and G. Alenyà. Causal explanations for robot decisions and beliefs, 2026 ACM/IEEE International Conference on Human-Robot Interaction, 2026, Edinburgh, Scotland, UK, pp. 1342-1344, ACM/IEEE.
Follow us!