H‑AI TEAM

PEPR eNSEMBLE · France 2030 · 2026–2030

The project

Human–AI Teams, Embedded Agency and Morality: philosophy and aesthetics of the collaboration with machines

H-AI TEAM is a four-year research project (September 2026 – August 2030) funded by the French National Research Agency under the France 2030 programme, within the PEPR eNSEMBLE devoted to digital collaboration. It brings together the Institute for the History and Philosophy of Science and Technology (IHPST), the Institute of Intelligent Systems and Robotics (ISIR, Sorbonne Center for Artificial Intelligence) and THALIM (Theory and History of Modern Arts and Literatures).

The problem

The AI systems currently being developed, and especially those now called agents, profoundly alter the social and epistemic environments in which we act. They do not merely mediate our information: they participate in it. Hence a growing pressure to “collaborate” with them and to include them, as autonomous agents, within human teams. Yet little is known about the effects of such collaborations on human autonomy, agency and creativity.

The project asks three questions:

  1. What kind of agency do AI systems display in our interactions with them?
  2. To what extent do these interactions affect our own agency?
  3. How do they affect the agency of the human team they join?

The approach

Rather than treating human–AI collaboration as self-evident, H-AI TEAM builds an articulated account of it along three axes: agency (what humans and machines can do), collaboration (how collective agency forms — or unravels) and the environment (physical, digital, hybrid) in which such collaborations are always embedded. This attention to environmental embedding, informed by the philosophy of biology and the notion of scaffolding, sets the project apart from existing approaches.

The work unfolds in five work packages: the foundations of agency (WP1), collective action (WP2), overreliance and human autonomy (WP3), creative collaboration in the arts (WP4) and design and governance recommendations (WP5). The study of AI-assisted artistic practices — with four French artists who have developed original co-creation protocols — serves as a test case: it puts philosophical models of distributed agency to a concrete test and sheds light on their social and legal stakes.

Expected outcomes

Conceptual tools to categorise our interactions, joint actions and teamwork with AI systems; an analysis of how individual and collective agency is affected by these systems; a typology of forms of dependence and indicators for detecting them; a genealogy and theory of co-agency in creation; and finally a white paper of design and governance recommendations for developers and regulators alike, in connection with the European framework (AI Act, DSA).

Consortium

  • IHPST (CNRS / Université Paris 1 Panthéon-Sorbonne) — coordination
  • ISIR, Sorbonne Center for Artificial Intelligence (CNRS / Sorbonne Université)
  • THALIM (CNRS / Université Sorbonne Nouvelle / École normale supérieure)

The project builds on the ANR CulturIA project (2021–2025), devoted to a cultural history of artificial intelligence, which notably led to the exhibition The World According to AI at the Jeu de Paume (2025).