27TH EUROPEAN CONFERENCE ON ARTIFICIAL INTELLIGENCE

Intelligent Sharing of Explanation Experience by Users for Users: Case-Based Reasoning for Explanation Strategy Reuse

Abstract: Different stakeholders of AI applications have diverse backgrounds, competencies, and goals, necessitating various types of explanations. Even individual users may require multiple explanations, and the same user might need alternative explanations as their mental models evolve. Additionally, the same explanation may need to be presented in different modalities. How can we determine the most effective methods or combinations of methods for interpreting and explaining ML models to meet the diverse needs of different AI models and users? The iSee project is addressing this question by employing a combination of explainers to form what we refer to as an "explanation strategy," enabling personalised and interactive explanations. Using Case-Based Reasoning (CBR) as the underpinning methodology, the project captures and reuses experiences of best practices in explanation strategies through the iSee platform. The talk will present the key components that facilitate reasoning in iSee: an ontology to model experiences, cases to capture these experiences, a retrieval engine to identify best practices supported by iSee's explanation evaluation questionnaire, and an interactive interface to engage with end-users.

Bio: Nirmalie Wiratunga is Professor of Intelligent Systems and Associate Dean for Research at the Robert Gordon University’s School of Computing in Aberdeen, Scotland. With over two decades of experience in AI and Reasoning research, she leads the Artificial Intelligence & Reasoning Research Group (AIR) and has been involved in numerous UK- and EU-funded projects, including platforms for reusable explainable AI experiences and healthcare initiatives. She is best known for her work on case-based reasoning with unstructured data, having won Best Paper Awards for her textual case representation and solution generation methods.