iToBoS XAI Workshop

The iToBoS consortium conducted an explainable AI (xAI) workshop at the project’s 6th General Assembly meeting, hosted by Bosch Manufacturing Solutions facility, in Madrid in October 2023.

In the project, artificial intelligence development continues to gather pace, and so workshop moderator and work package 2 lead, Trilateral Research felt the gathering of consortium members was a perfect opportunity to initiate discussion about important explainable and ethical AI topics.

The consortium members first provided an overview of the current status of AI development and implementation within their corresponding work packages and the presentations focused on three strands of research:

  1. Lesion detection, segmentation, and classification, presented by Torus Actions;

  2. Development of the risk profiling using tabular data, presented by NTUA;

  3. An overview of the concept of ethical AI, presented by Trilateral Research.

After the presentations, Trilateral Research moderated three breakout group discussions. The breakout groups were organised so that representatives from each of the primary stakeholder groups were included in each of the discussions. Within the project, the primary stakeholder groups are researchers, clinicians, and patient advocates.

The breakout groups were organised along the following themes and research questions:

  1. Which explainability requirements do each of the stakeholder groups (clinicians, patients, researchers) have?

  2.  What do clinicians and patients need to know about the AI models inner working to be able to trust them?

  3.  How do you think explainability and ethical AI should be embedded in the development lifecycle?

The breakout group discussions are viewed as a critical initiation for further elaboration by the consortium, with the goal of achieving some of the aims of the iToBoS project. It was also used to understand different perspectives both inside and outside of the consortium, with participants being able to talk about their domain expertise and knowledge of AI locally and globally. The outcome of the discussions will feed into the work being led by Trilateral Research, within work package 2. More specifically the conducting of the privacy, social and ethical impact assessment, which will try and outline a number of risks, their probability of occurring, and proposal of mitigations.