Programme

Programme détaillé

Session 2 (Chair TBD)

• MIIC-SR: From Complex Data to Structural Causal Models (Nadir Sella, Arefe Asadi, Myriam Tami et Louis Verny)

• Perspectives for Direct Interpretability in Multi-Agent Deep Reinforcement Learning (Yoann Poupart, Aurélie Beynier et Nicolas Maudet)

• Vers une couverture partielle d'explications de réseaux neuronaux par des réseaux d'approximation (Mathieu Brassart et Laurent Simon)

Conférencière invitée: Katrien Verbert

Empowering users through interactive and hybrid explanations

Résumé

Despite the long history of work on explanations in the Machine Learning, AI and Recommender Systems literature, current efforts face unprecedented difficulties: contemporary models are more complex and less interpretable than ever. As such models are used in many day-to-day applications, justifying their decisions to end users will only become more crucial. In addition, several researchers have voiced the need for interaction with explanations as a core requirement to support empowerment of users. Such interaction methods can enable users to steer models with input and feedback, and can support better model understanding.  In this talk, I will present our work on interactive explanation methods tailored to the needs of end users, such as healthcare professionals and job seekers. In addition, I will present our work on combining data-centric and model-centric explanations to empower end users in refining predictive models.  Our work emphasizes interactive, hybrid explanation methods that not only improve model understanding but also enhance the user’s ability to steer and improve AI models using domain knowledge. 

Session 3 (Chair TBD)

X-Train: eXplanations for Training – Utilisation d'explications pour l’entraînement Multimodal des Transformers (Meghna P Ayyar, Jenny Benois-Pineau and Akka Zemmari)

Attention and Beyond: Explainability Techniques for Vision Transformers (Wadie El Amrani)

Table ronde : slides