From Specification Models to Explanation Models: An Extraction and Refinement Process for Timed Automata

TL;DR

This paper presents a process to extract and refine explanation models from system models, enabling autonomous systems to generate context-specific explanations and improve trust and safety.

Contribution

It introduces a high-level process for deriving explanation models from specification models, facilitating reuse and customization for autonomous systems.

Findings

Effective extraction of explanation models demonstrated with autonomous driving example

Refinement process allows tailoring explanations to users and situations

Enhances trust and safety in autonomous systems through better explanations

Abstract

Autonomous systems control many tasks in our daily lives. To increase trust in those systems and safety of the interaction between humans and autonomous systems, the system behaviour and reasons for autonomous decision should be explained to users, experts and public authorities. One way to provide such explanations is to use behavioural models to generate context- and user-specific explanations at run-time. However, this comes at the cost of higher modelling effort as additional models need to be constructed. In this paper, we propose a high-level process to extract such explanation models from system models, and to subsequently refine these towards specific users, explanation purposes and situations. By this, we enable the reuse of specification models for integrating self-explanation capabilities into systems. We showcase our approach using a running example from the autonomous driving domain.