Scalable Verification of Social Explainable AI by Variable Abstraction

Wojciech Jamroga, Wojciech Jamroga, Yan Kim, Damian Kurpiewski

2024

Abstract

Social Explainable AI (SAI) is a new direction in artificial intelligence that emphasises decentralisation, trans-parency, social context, and focus on the human users. SAI research is still at an early stage, and concentrates mainly on delivering the intended functionalities. At the same time, formal analysis and verification of the proposed solutions is rare. In this paper, we present an approach to formal verification of SAI protocols by means of temporal model checking. We use agent graphs to model benign as well as malicious behaviors of the participants, branching-time logic formulas to express interesting properties of the protocol, and the state of the art temporal model checker UPPAAL to verify those formulas. As usual in such cases, state-space explosion and the resulting complexity of verification is a major problem. We show how to mitigate the complexity through state abstraction, and demonstrate the advantages in practice by using a novel tool for user-friendly abstractions EASYABSTRACT4UPPAAL.

Paper Citation

in Harvard Style

Jamroga W., Kim Y. and Kurpiewski D. (2024). Scalable Verification of Social Explainable AI by Variable Abstraction. In Proceedings of the 16th International Conference on Agents and Artificial Intelligence - Volume 1: ICAART; ISBN 978-989-758-680-4, SciTePress, pages 149-158. DOI: 10.5220/0012474800003636

in Bibtex Style

@conference{icaart24,
author={Wojciech Jamroga and Yan Kim and Damian Kurpiewski},
title={Scalable Verification of Social Explainable AI by Variable Abstraction},
booktitle={Proceedings of the 16th International Conference on Agents and Artificial Intelligence - Volume 1: ICAART},
year={2024},
pages={149-158},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012474800003636},
isbn={978-989-758-680-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 16th International Conference on Agents and Artificial Intelligence - Volume 1: ICAART
TI - Scalable Verification of Social Explainable AI by Variable Abstraction
SN - 978-989-758-680-4
AU - Jamroga W.
AU - Kim Y.
AU - Kurpiewski D.
PY - 2024
SP - 149
EP - 158
DO - 10.5220/0012474800003636
PB - SciTePress