Exploring the role of embodiment on intimacy perception in a multiparty collaborative task
New research shows how an AI agent's physical form changes group dynamics and perceived closeness.
A team of researchers from the Laboratoire Interdisciplinaire des Sciences du Numérique (LISN) and LITIS has published a new study investigating a critical factor in human-AI collaboration: embodiment. The paper, 'Exploring the role of embodiment on intimacy perception in a multiparty collaborative task,' examines how the physical or virtual form of an AI agent influences a group's sense of cohesion and intimacy during a shared activity. The researchers built their analysis on a collected corpus of human players engaging in a collaborative board game, using this data to study how different embodiments affect the perception of the agents and the group's overall dynamic.
The core finding is that an agent's embodiment is a significant variable in shaping social interactions within a team, impacting everything from task success to the development of interpersonal relationships. The study moves beyond pure task efficiency to focus on the social skills necessary for effective multi-entity collaboration. The researchers conclude by outlining the key problematics and challenges involved in designing experimental protocols and functional multi-agent systems that can authentically replicate or enhance human-like group cohesion, with their work accepted for the ACM International Conference on Intelligent Virtual Agents (IVA) in 2024.
- Study by LISN/LITIS uses human board game corpus to analyze AI embodiment's social impact.
- Finds agent's physical/virtual form significantly affects group intimacy perception and cohesion.
- Outlines design challenges for creating collaborative multi-agent systems, presented at ACM IVA '24.
Why It Matters
This research is crucial for designing AI teammates and virtual assistants that can build trust and work seamlessly within human teams.