The quest of parsimonious XAI: A human-agent architecture for explanation formulation

With the widespread use of Artificial Intelligence (AI), understanding behavior intelligent agents and robots is crucial to guarantee successful human-agent collaboration since it not straightforward for humans understand an agent's state mind. Recent empirical studies have confirmed that explaining a system's human users fosters latter's acceptance system. However, providing overwhelming or unnecessary information may also confuse cause failure. For these reasons, parsimony has been outlined as one key features allowing interaction with parsimonious explanation defined simplest ( i.e. least complex) describes situation adequately descriptive adequacy). While receiving growing attention in literature, most works are carried out on conceptual front. This paper proposes mechanism eXplainable AI (XAI). In particular, introduces process formulation HAExA, explainability architecture make operational remote robots. To provide explanations, HAExA relies both contrastive explanations filtering. evaluate proposed architecture, several research hypotheses investigated user study well-established XAI metrics estimate how trustworthy satisfactory provided by are. The results analyzed using parametric non-parametric statistical testing.