Woman reading a note against a background that represents AI

Insight Excellence: Explainable AI for Transparent and Trustworthy Decision Making

Submitted on Wednesday, 17/05/2023

The importance of interpretable and trusted explanations of AI decision making is widely recognised as critical for the future acceptance of AI. Insight is an international leader in this field, mixing foundational computer science with interdisciplinary work in cognitive science, in applications and in informing national and international policy.

We have continued to make significant advances in 2022, with our work that focuses on Counterfactual Explanations (Dai et al, 2022)(Temraz et al 2022)(Kenny and Keane 2021). We have concentrated primarily on making automated explanations understandable to end users, developing new models based on an understanding of cognitive processes. Complementary work investigated producing textual descriptions from deep learning models using knowledge graphs for tasks such as image classification (dos Santos et al 2023)(Horta and Mileo, 2022). Explainable recommendations in recommender systems is another important application domain targeted (Wang et al, 2022), including developing novel techniques for identifying and quantifying algorithmic bias, particularly in multi-modal contexts (Mandal et al 2023). All of these activities are collated and discussed by the national network mobilised through Insight’s Trustworthy AI PRI.

In May 2022, Barry O’Sullivan was appointed by Minister for Trade Promotion, Digital and Company Regulation, Robert Troy TD, to the Enterprise Digital Advisory Forum, an advisory group to the Government on matters of AI and digital policy. Professor O’Sullivan also chaired the Expert Group for Future Skills Needs Committee on Artificial Intelligence Skills. The final report of this committee was published in June 2022 by Minister for Trade Promotion, Digital and Company Regulation, Robert Troy TD, and Minister for Skills and Further Education, Niall Collins TD. Finally, Professor O’Sullivan has been working with the Department of Enterprise, Trade, and Employment on the creation of a cross-departmental group focus on developing a Trustworthy AI Charter for use across Government to inform all public sector uses of AI. This group was formally established in the Autumn of 2022 under the chairing of Barry Lowry, Government Chief Information Officer (OGCIO), and Jean Carbery, Assistant Secretary of the EU, Digital and Access to Finance Division (DETE).

Investigators: Rouhai Dong (UCD), Mark Keane (UCD), Suzanne Little (DCU), Alessandra Mileo (DCU), Barry O’Sullivan (UCC), Susan Leavy (UCD), Aonghus Lawlor (UCD), Neil Hurley (UCD), Paolo Palmieri (UCC), Barry Smyth (UCD)