Towards Balancing Preference and Performance through Adaptive Personalized Explainability

As robots and digital assistants are deployed in the real world, these agents must be able to communicate their decision-making criteria to build trust, improve human-robot teaming, and enable collaboration. While the field of explainable artificial intelligence (xAI) has made great strides in building a set of mechanisms to enable such communication, these advances often assume that one approach is ideally suited to each problem (e.g., decision trees for explaining how to triage patients in an emergency or feature-importance maps for explaining radiology reports). This fails to recognize that users may have different experiences or preferences for interaction modalities. In this work, we present the design and results of two user-studies set in a simulated autonomous vehicle (AV) domain, a setting that is increasingly important to HRI. We investigate (1) population-level preferences for xAI and (2) different personalization strategies for providing robot explanations. We find significant differences between xAI modes in both preference (p < 0.01) and task-performance (p < 0.05). We also observe that a participant's preferences do not always align with their task-performance, motivating our development of an adaptive personalization strategy that balances the two. We show that this strategy leads to significant performance gains (p < 0.05), and we conclude with a discussion our findings and implications for future work in xAI.