Résumé
In a world where increasingly important decisions are made by computational systems, the trustworthiness of Artificial Intelligence (AI) has become a central concern. To be trustworthy, an AI system must be lawful, ethical, and robust, allowing human agency and oversight [2]. It must also ensure user well-being and privacy, as well as providing transparency, safety, fairness, and accountability. High-stakes scenarios, such as health applications, that involve potentially high-risk and profoundly impactful decisions underscore the importance of these trustworthy requirements. In our respective works, we tackle the robustness aspect of trustworthy AI through uncertainty quantification, aiming to inform decision-makers about the confidence level of AI-generated results. The idea is to be able to give, to the decision-makers, an information about the uncertainty of the given result. This uncertainty could come both from the data and the AI model itself [6]. For example, it could be a subset of classes for a classification task or an interval for a regression one. In this doctoral consortium presentation, we introduce two different projects that apply Trustworthy AI methods to movement-related areas.