This piece originally appeared in Tech Policy Press.
Many in the tech industry expect that in the near term AI will evolve from the level of today's chatbots to more sophisticated AI agents capable of engaging in complex actions with little supervision. First, this raises the prospect that democratic freedoms will increasingly be threatened by the failure to adequately consider the essential role of personal agency in how our agents serve us. Second, beyond effects on individual agency and autonomy, lies the potential role of AI agents in the broader community and effects on the social dynamics that our humanity is built on. We suggest that new and stronger relationship framings are needed to ensure that AI serves humanity.
Our thesis is that the growing body of work on “AI ethics,” “value alignment,” “safety,” and “guardrails” raises a host of important issues, but needs to spotlight a strategic focus on the fundamental core of the threats to democracy and human freedom. These arise from how AI will increasingly shape what we know of the world, how we act on the world, and how AI acts for us. This piece synthesizes work by the two of us to propose novel governance framings along with technical approaches to help identify and remedy those challenges to human choice and autonomy.
The overarching idea presented below is that just as humanity is not monolithic but contains multitudes, our agents or assistants cannot be built as monolithic systems that are somehow aligned by their developers to serve all interests. Rather, these systems must be composed of interoperating agents that faithfully serve their users and, in doing so, negotiate a broader alignment. This must be done at both the individual and community levels. For every broadly functional AI agent or assistant system component, the essential question is, “whom does it serve?”
Further, it now seems likely there will be a dramatic realignment of US AI policy development, with reduced emphasis on regulatory guardrails – and that would have global impact. We believe the framings and strategies suggested below offer a foundation in individual agency, leavened by social mediation, that will organically enable the adaptability, resilience, and collective intelligence needed to advance human flourishing, even in times of polarization, turbulence and governmental gridlock. Thus, we suggest that solutions that build on these framings and strategies should be able to garner broad bipartisan support once these principles are widely recognized.
