Help shape how AI learns to work with people.
Whether you're researching personalized agents, evaluating human–AI collaboration, or building new interactive systems, we'd love to hear your ideas. No standing committee membership or recurring duties required.
Research that keeps people in the loop.
Learning to Work with You is not just about more capable systems. It is about building AI that collaborates in ways that respect human agency, diverse needs, and different preferences for oversight.
As the community grows, this space can support shared questions, reading groups, benchmarks, and collaborations across personalization, mixed-initiative interaction, and human–AI collaboration.
Research updates
A mailing list is planned. For now, you can express interest in future updates and research conversations.
Express interest →Contribute literature
Suggest papers, improve labels, and refine the research taxonomy through reviewed GitHub changes.
Research map GitHub ↗Share an idea
Tell us about an open problem, benchmark design, missing research area, or potential collaboration.
Send an idea →A starting point for joint research.
These are invitations for discussion—not staffed projects or existing partnerships.
Learning from sparse procedural feedback
Can just a few corrections or approval decisions help predict future oversight preferences?
Discuss this question →Context-conditioned autonomy
Can an agent adapt when the same person wants more confirmation in one situation than another?
Discuss this question →Benchmarking USIL
How can we measure unwanted autonomy, unnecessary interruptions, and longitudinal transfer?
Discuss this question →Start with a conversation.
This form prepares a message in your email application; it does not sign you up to a mailing list or store your information on this website.
Your message goes to the founding research lead. Your name or comments won't be published as a community endorsement without your permission.
Community organizers, advisors, and institutional partners will be listed only after their roles and permissions are confirmed.