Hierarchical Expertise Level Modeling for User Specific Contrastive Explanations

Abstract

There is a growing interest within the AI research community in developing autonomous systems capable of explaining their behavior to users. However, the problem of computing explanations for users of different levels of expertise has received little research attention. We propose an approach for addressing this problem by representing the user's understanding of the task as an abstraction of the domain model that the planner uses. We present algorithms for generating minimal explanations in cases where this abstract human model is not known. We reduce the problem of generating an explanation to a search over the space of abstract models and show that while the complete problem is NP-hard, a greedy algorithm can provide good approximations of the optimal solution. We also empirically show that our approach can efficiently compute explanations for a variety of problems.

The challenge of crafting intelligible intelligence

“…Depending on the choice of explanatory foil, different answers are appropriate [8]. Sreedharan et al describe an algorithm for generating the minimal explanation that patches a user's partial understanding of a domain [37]. Work on mixed-initiative planning [7] has demonstrated the importance of supporting interactive dialog with a planning system.…”

A Survey of Contrastive and Counterfactual Explanation Generation Methods for Explainable Artificial Intelligence

“…The problem of obtaining contrastive explanations is designed as a Bayesian inference problem, with the posterior distribution to be maximized defined as the probability of a contrastive explanation given a set of positive and negative plan traces. Conversely, Sreedharan et al consider the task of automatic analysis of counterfactual explanations in their ''Hierarchical Expertise-Level Modeling'' framework [156]. A robot provides a user with a plan for the next action to take.…”

“…Contfactual explanation generation appears highly relevant to sequential tasks in robotics such as automatic planning [94], [157], [161]. Moreover, some of the robotics-related frameworks found in reinforcement learning settings provide explanations for policies that a robot selects at a given time step [132], [156].…”

The Emerging Landscape of Explainable Automated Planning & Decision Making

“…Contrastive. The contrastive nature of these explanations comes from how the model update preserves τ of the given plan as opposed to the foil, which may be implicitly [15] or explicitly [72] provided. This is also closely tied with the selection process of those model updates.…”

“…In [71], authors show how to reconcile with a set of possible mental models {Π H i } and also demonstrate how the same framework can be used to explain to multiple users in the loop. In [72], on the other hand, the authors estimate the mental model from the provided foil.…”