• Classified by Topic • Classified by Publication Type • Sorted by Date • Sorted by First Author Last Name • Classified by Funding Source •
Coadaptive Value Alignment.
Nathan Tsoi, Eric Hsiung, Masayuki Yamazaki, Swarat Chaudhuri, Peter
Stone, and and Joydeep Biswas.
In Proceedings of the 25th International
Conference on Autonomous Agents and Multiagent Systems, January 2026.
The integration of autonomous agents into human society is a grand challenge forAI. In order to achieve widespread acceptance, agents must conform to the valuesof people with whom they interact. Current approaches treat the value alignmentproblem as a unidirectional interaction where the aim is to imbue an agent'sactions with human values. Our Coadaptive Value Alignment paradigm acknowledgesthat human perceptions, expectations, and values continuously evolve in responseto agent actions. We conceptualize human-agent interaction as an adaptive loopwhere the agent actively models and intentionally influences the human'sperception, rather than just acting according to static human values. Forinstance, unlike a traditional agent that simply maximizes speed, an adaptiveagent could detect a drop in user trust and strategically sacrifice short-termefficiency to repair the relationship. This perspective transforms valuealignment into a multi-agent challenge where all actors must identify and adhereto a shared, implicit social contract. The opportunity to create a virtuous cycleof self-improvement is accompanied by the risk of negative reinforcement, whichcould result in undesired behaviors. We outline the core framework components,present a research road map for the MAS community, and propose that this dynamicperspective is critical for creating truly collaborative social partners.
@InProceedings{tsoi2026coadaptive,
author = {Nathan Tsoi and Eric Hsiung and Masayuki Yamazaki and Swarat Chaudhuri and Peter Stone and and Joydeep Biswas},
title = {Coadaptive Value Alignment},
booktitle = {Proceedings of the 25th International Conference on Autonomous Agents and Multiagent Systems},
year = {2026},
month = {January},
location = {Paphos, Cyprus},
abstract = {The integration of autonomous agents into human society is a grand challenge for
AI. In order to achieve widespread acceptance, agents must conform to the values
of people with whom they interact. Current approaches treat the value alignment
problem as a unidirectional interaction where the aim is to imbue an agent's
actions with human values. Our Coadaptive Value Alignment paradigm acknowledges
that human perceptions, expectations, and values continuously evolve in response
to agent actions. We conceptualize human-agent interaction as an adaptive loop
where the agent actively models and intentionally influences the human's
perception, rather than just acting according to static human values. For
instance, unlike a traditional agent that simply maximizes speed, an adaptive
agent could detect a drop in user trust and strategically sacrifice short-term
efficiency to repair the relationship. This perspective transforms value
alignment into a multi-agent challenge where all actors must identify and adhere
to a shared, implicit social contract. The opportunity to create a virtuous cycle
of self-improvement is accompanied by the risk of negative reinforcement, which
could result in undesired behaviors. We outline the core framework components,
present a research road map for the MAS community, and propose that this dynamic
perspective is critical for creating truly collaborative social partners.
},
}
Generated by bib2html.pl (written by Patrick Riley ) on Mon May 18, 2026 15:31:22