Observing, understanding, accompanying: the STARS team’s research dynamics in the era of augmented human interaction

Its aim is to better understand social behavior in its real-life context, through the observation of gestures, voices, eye contact and physiological signals.

Research focuses on several complementary areas:

• Analysis of social interactions, with particular attention to group dynamics, non-verbal gestures and the distribution of speech;
• Support for mental health diagnosis, through clinical collaborations aimed at detecting disorders such as depression, schizophrenia or post-traumatic stress;
• Stress and emotion recognition, via multimodal datasets and wearable sensors;
• Video anomaly detection (falls, isolation, risky gestures), with weakly supervised approaches;
• Multi-object tracking, combining boxes and masks to refine the analysis ;
• Video Generation: Leveraging Generative Models for Data Augmentation and Anonymization.

Behind all this work lies the same desire: to design tools that are precise, yet respectful of people, always designed to support rather than monitor.

One of the team’s future projects is Créapolis, a daytime activity center in Sophia Antipolis, designed to stimulate the cognitive and social abilities of the elderly.

The STARS team sees it as a concrete testing ground for some of its systems currently under development, in a lively, non-medical setting. Cognitive games, filmed interactions, workshops based on artificial intelligence: several ideas are already ready, awaiting ethical and logistical validation.

Two startups led by former doctoral students could find their first application here:

• Facila, focused on voice training;
• ThinkSync, a conversational assistant designed to help structure thoughts.

While nothing has yet been launched on site, discussions are well underway. The presence of a dedicated video room in the building under construction also opens up prospects for respectful, targeted observation.

These exchanges are being built up thanks in particular to a key contact: Philippe Robert, Director of Alzheimer’s Innovation and one of the people in charge of Créapolis. He’s also clinical advisor for the Mephesto project, which focuses on understanding psychiatric illnesses associated with aging, schizophrenia, post-traumatic stress disorder and bipolar.

Through him, team members have already been able to visit the premises, observe certain activities, and discuss possible applications. The link is a natural one: Mephesto is interested in pathologies such as Alzheimer’s, schizophrenia and bipolar disorders; STARS, for its part, is working on tools to better characterize them.

At the same time, the team is developing a clinical support system for mental health professionals. Based on observable signs - pauses, gestures, eye contact - the aim is to generate indicators that complement conventional assessment, to help identify symptoms.

At Créapolis, there is no question of psychiatric assessment. The team wants to stick to observing human interactions, without any clinical framework: for example, two people playing cards, and we analyze their gestures or the rhythm of their exchanges. The idea is to remain within a logic of behavioral interpretation, without a diagnostic approach.