Towards Modelling and Verification of Social Explainable AI

TL;DR

This paper introduces a formal approach to model and verify Social Explainable AI (SAI), focusing on its resilience against malicious activities using strategic ability logics and the STV model checker.

Contribution

It pioneers the application of multi-agent strategic ability models and formal verification techniques to assess SAI's robustness against attacks.

Findings

First formal modeling of SAI environments

Verification of resistance to compromised AI modules

Use of STV model checker for strategic ability analysis

Abstract

Social Explainable AI (SAI) is a new direction in artificial intelligence that emphasises decentralisation, transparency, social context, and focus on the human users. SAI research is still at an early stage. Consequently, it concentrates on delivering the intended functionalities, but largely ignores the possibility of unwelcome behaviours due to malicious or erroneous activity. We propose that, in order to capture the breadth of relevant aspects, one can use models and logics of strategic ability, that have been developed in multi-agent systems. Using the STV model checker, we take the first step towards the formal modelling and verification of SAI environments, in particular of their resistance to various types of attacks by compromised AI modules.