Towards a User-Centered, Goal-Driven, and Context-Involved Explaining System
Keywords: XAI
Abstract: In this essay, we focus on characteristics of selected and social in human explanations, thus arguing that, a user-centered, context-involved and goal-driven system is what we need to build a good explaining system. Two characteristics of human explanations, selected, i.e., select the most relevant as explanations, and social, i.e., how to select is determined by the context of the targeted users, are first introduced based on psychological research. Then, three forms of human explanations, deductive proofs, casual patterns, and mental models, are analysed about their advantages and limitations from the view of selected and social explanations. The consequent challenge is that all these theories are usually without context surrounding and not universally applicable. Finally, two ways of designing priors for types of users and applying bidirectional human-robot communication models are proposed to offer possible avenues.
Submission Number: 210
Loading