Embodied AI & Decision Support
Our core research: AI with a face or a body that helps people decide while the decision stays theirs. How avatars, digital humans and digital teammates perform, how people trust and respond to them, and how to make them safe, accessible and honest.
2 entries
- VideoFDB: Evaluating Full-Duplex Vision-Speech Capabilities in Conversational Agents
A benchmark that tests whether AI agents can listen, speak and show nonverbal signals at the same time, as people do on a video call. It finds today's systems fall well short of humans, and that avatar pipelines built from separate speech and video steps cannot react while the user is still talking.
- A multidimensional measurement of photorealistic avatar quality of experience
A study that measures how people experience photorealistic avatars on ten human dimensions, including realism, trust, comfort and creepiness, using an open crowdsourced test. It finds that the more realistic an avatar is, the more people like and trust it, with no uncanny-valley dip in video-call settings.
Also see: AI Policy & Industrial Strategy