Investigating and measuring the effect of cyclic relationship on user-XAI interaction

Abstract

Establishing a cyclic and symbiotic relationship between human and artificial intelligence is essential part of ensuring success of explanatory interfaces. However, despite the development in the field of explainable artificial intelligence systems, a concrete quantitative measure for evaluating the usability of such systems is nonexistent. We, therefore, propose explanatory efficacy, a novel metric to evaluate the strength of the cyclic relationship the interface exhibits. In addition, in a user study, we analyzed participants' affective-cognitive processes and recorded their EEG signals as they interacted with our custom built personalized recommendation system. We found that system with high explanatory efficacy affects both system's performance and affective-cognitive state of users positively. Furthermore, our results indicate feasibility of EEG as a measure of explanatory efficacy.