Generating Explanations for Temporal Logic Planner Decisions

Abstract

Although temporal logic has been touted as a fruitful language for specifying interpretable agent objectives, there has been little emphasis on generating explanations for agents with temporal logic objectives. In this paper, we develop an approach to generating explanations for the behavior of agents planning with several temporal logic objectives. We focus on agents operating in deterministic Markov decision processes (MDPs), and specify objectives using linear temporal logic (LTL). Given an agent planning to maximally satisfy some set of LTL objectives (with an associated preference structure) in a deterministic MDP, we introduce an algorithm for constructing explanations answering both factual and “why” queries, which queries are also specified in LTL.

Explaining Autonomous Decisions in Swarms of Human-on-the-Loop Small Unmanned Aerial Systems

“…The explainability of intelligent systems and the impact of explanations upon the user's mental model have been widely studied in various contexts including machine learning classifiers & AI models (Chakraborti et al 2017), recommendation systems (Dominguez et al 2019), and planning & decision systems (Kasenberg, Thielstrom, and Scheutz 2020;Zakershahrak et al 2020). In the context of autonomous agents, Titarev et.al (Tintarev and Kutlak 2014) developed Scrutable Autonomous Systems (SAsSy) to demonstrate a humanunderstandable dialog explaining an autonomous agent's behavior through argumentation and natural language.…”