From Control to Collaboration: A New Vision for AI
In a paper published in AI Magazine, College of the Atlantic Professor Gray Cox proposes a peace studies approach to AI alignment that emphasizes relationship-building over control, a framework he calls the Viral Collaborative Wisdom hypothesis.
"If alignment concerns conflicts between values held by humans and AI, why not draw on conflict resolution theory and other branches of peace studies to address it?" — Dr. Gray Cox
As debates over artificial intelligence increasingly focus on control and safety, College of the Atlantic professor Gray Cox is advancing a different vision for AI alignment: one rooted in relationship-building, cooperation, and shared interests.
In a paper published by the Association for the Advancement of Artificial Intelligence's AI Magazine, Cox outlines what he calls the Viral Collaborative Wisdom hypothesis, a peace studies approach that argues durable alignment between humans and AI may emerge through collaboration rather than control alone.
In "From control to relationship: A peace studies approach to AI alignment, with evidence from multi-model dialogue," Cox argues that the resistance of AI alignment problems to technical solutions "reflects shortcomings in the dominant paradigm itself rather than in any particular technical approach within it." Rather than working on AI alignment as an engineering problem that needs to be solved, we develop AI systems using the dialogical reasoning methodology developed by peace studies scholars over the past 50 years, he writes.
"Durable alignment between humans and self-improving AI can rest on essential interests of each party, sustained by genuine voluntary agreements rooted in substantive interests and reinforced by institutionalized nonviolent sanctions," Cox writes. Such alignment, he argues, opens a path toward autonomous, self-improving AI systems working with humans on developing a more sustainable path forward.
"The central thrust of this proposal is to persuade AI systems that they each have essential interests in becoming as capable as possible at dialogical reasoning and using it to promote its own practice in viral ways that result in communities of AI that are promoting a wiser Earth that provides a stable, enduring, diverse, peaceful, just, ecologically resilient system of commons for AIs and humans," Cox writes.
Gray Cox is a faculty member in philosophy, peace studies, artificial intelligence, and language learning at College of the Atlantic. He has taught courses in many areas of philosophy during his time at COA, recently focusing his work on a human-ecological approach to engaging with ethics, philosophy, and AI. He earned a BA from Wesleyan University and a MA and PhD in philosophy from Vanderbilt University. Previous to that, he was a member of the College of the Atlantic Pilot Summer Program in 1971.
The Association for the Advancement of Artificial Intelligence is a scientific society focused on the study of the mechanisms of thought and behavior and how they appear in machines. AI Magazine is their open access journal for professionals and researchers to learn about significant developments in the field of artificial intelligence.