iSee (isee4xai.com) is a CHIST-ERA funded project being developed by a consortium of European universities (Complutense University of Madrid, Robert Gordon University, University College Cork) and an industrial partner (BT France). It aims to build a platform that captures, stores and re-uses end-user explanation experiences for various AI models using Case-Based reasoning methodology.
Cases are formed of knowledge of the AI model and its user group (problem component), the explanation strategy recommended (solution component), and feedback from the user group to describe whether the provisioned explanations were satisfactory (outcome component). In this manner, cases represent a comprehensive record of explanation experience.
In this tutorial, the iSee project team would like to share with the participants the latest developments of this project, especially:
- How explanation experiences are represented in an ontology created for the project
- How explanations experiences are captured in a case-based database
- How use cases can be designed and represented in the system
- How explanations are progressively and interactively provided to end-users using a chatbot and how they can be evaluated.
- How end-user evaluations are measured and used to augment the explanations strategies case base.
The tutorial will culminate in attendees being able to add a ‘mock’ case to the system, and receive recommendations of an explanation strategy to suit that case. Pending interest from the academic community, the tutorial may be expanded to include an interactive component demonstrating the uploading of XAI algorithms to iSee’s explainer API for broader industry impact.
Interest to the ICCBR attendees
This tutorial session would enable the audience to get a better understanding of the recent developments, provide feedback that can be considered for inclusion in the roadmap, and get an opportunity to submit their own contributions to expand the visibility and reach of their work. The tutorial would also be a great real-life example on how Case-Based reasoning approaches can be utilized in the context of Explainable AI, and how the iSee project proposes a novel
contribution in this field.
Beyond these reasons, iSee will launch as an explanation platform with CBR at its heart next year. We envision that iSee will be a tool that offers avenues for researchers to broaden and deepen their research impact, so having an early understanding of the tool (as well as the ability to influence its development) is beneficial for CBR researchers. On top of this, the tool offers many opportunities for researchers to get involved and update components; from design of
novel AI algorithms, to new evaluation methodologies, or expansions to the ontology. There are significant research avenues to explore.
Tutorial session format
This tutorial would be delivered as a 2-hour session that will cover the key points listed above. The session aims to be as much interactive as possible with live demos, constant exchanges with the audience and feedback times to capture new ideas and/or suggestions or improvements.
A real-life use case could be selected from the audience if possible as an example, or the organizers could choose one of the existing use cases from within the iSee project, including health (explanation of intelligent fracture diagnosis in x-rays), telecommunication (explanation of dynamic field engineer support) or assisted living (explanation of smart home predictions for tenants).
Ideally, this tutorial would occur in advance of the XCBR 2023 Workshop/Challenge, as it is likely to be of interest to that audience specifically.
Previous related workshops
A XCBR workshop, organized by this project team, was held during the last ICCBR conference in 2022. It has received multiple submissions (6-8 papers approx.) from different research groups. Previous XCBR workshops included an invited talk and oral presentations of accepted papers.
All former participants of the past XCBR workshops will be contacted and asked to join this tutorial session. The session will also be advertised on various online media, such as iSee website, iSee LinkedIn users group and mailing lists.
Anne Liret email@example.com Organiser
Bruno Fleisch firstname.lastname@example.org Organiser
Chamath Palihawadana email@example.com Organiser