Trusted AI

As AI becomes more and more ubiquitous and penetrates fields where it was still unknown until recently (human resources management, administrations, fintech, basic sciences, e-health, justice, industry 4.0, etc.), it becomes obvious that a relationship of trust must be established between users and AI. Whether they are experts or not, people confronted with AI are entitled to expect certain guarantees (reliability, respect for data confidentiality, stability and consistency of decisions, etc.).

Work package 2 “Trust mechanisms for AI” seeks to meet this need from different angles: federated learning based on blockchain technologies such as Trusted Coalitions for Blockchained Distributed Learning (TCLearn) and BFAs; inductive logic for predictive justice; guarantees such as stability, choice of good metaparameters (especially in deep learning), self-assessment and certification; robustness to goal variations, to the quality of supervision and to unseen data, including in internal representations; model interpretability via hybridisation, distillation, constraints and complexity-optimality trade-offs; interaction, for example for infovis and multi-agent systems for by-design guarantees for robots and swarms All these different angles will be studied under this research axis.

The work package is coordinated by Prof. Benoît Frénay and Dr. Rebecca Marion at UNamur. A call has been launched to hold a first kick-off meeting of WP2 in March, after the start of the second semester. The aim will be to identify expressions of interest and to clarify the tasks that ARIAC researchers are working on. Regular meetings will then be planned (e.g. every two months) to ensure cohesion and visibility on the progress of the different teams and also to have a view on how the research work and the sub-tasks of the Grand Challenges may fit together. A first workshop will be organised in the framework of the 33rd International Francophone Conference on Human-Computer Interaction (HCI’22) to be held at the University of Namur from 5 to 8 April 2022, Namur (https://ihm2022.afihm.org/fr/ ) with the theme “Human-Computer Interaction and Explicability in Artificial Intelligence”. The WP2 (and more widely ARIAC) speakers will be invited to contribute.

At the moment, a total of 20 researchers, 14 of whom are funded by the ARIAC project, are working on this research axis. We are pleased to have been able to attract expertise from several partners. Most of the ARIAC researchers are affiliated with UNamur, UCLouvain and CETIC.