Keywords: Self explanation, Intelligent agents, Explainable agency
TL;DR: This position paper analyzes different types of self explanation that can arise in planning and related systems.
Abstract: In this paper, I discuss some varieties of explanation that can arise
in intelligent agents. I distinguish between process accounts, which
address the detailed decisions made during heuristic search, and
preference accounts, which clarify the ordering of alternatives
independent of how they were generated. I also hypothesize
which types of users will appreciate which types of explanation.
In addition, I discuss three facets of multi-step decision making
-- conceptual inference, plan generation, and plan execution --
in which explanations can arise. I also consider alternative ways
to present questions to agents and for them provide their answers.
Author Identity Visibility: Reveal author identities to reviewers
5 Replies
Loading