Formal Explainability for Neuro-Symbolic Artificial Intelligence. Artificial Intelligence (AI) is widely used in decision making procedures in many real-world applications, but weaknesses in the reaso

Description

Formal Explainability for Neuro-Symbolic Artificial Intelligence. Artificial Intelligence (AI) is widely used in decision making procedures in many real-world applications, but weaknesses in the reasoning capability of black-box AI has led to the development of neuro-symbolic AI combining the strength of black-box learning with reasoning. This project aims to develop methods to formally reason about and explain decisions of neuro-symbolic AI systems. Expected outcomes of this project are effective methods to explain to humans why a neuro-symbolic AI system makes a certain decision, using formal methods so that explanations are guaranteed to be correct. This should provide significant benefit since widespread use of neuro-symbolic AI will require the trust engendered through explainabilty. . Scheme: Discovery Projects. Field: 4602 - Artificial Intelligence. Lead: Prof Peter Stuckey