Political AI with a cooperative mandate

The problem

Most policy is written with a focus on pleasing the ruling party's electoral base or donors, resulting in ineffective laws that get overturned as soon as the balance of power shifts. It's also often written as a zero-sum exercise, with a focus on getting what's wanted by taking it from others or limiting them, instead of creating it.

The approach

Each Policy Lab debate assigns AI agents to represent real political constituencies: groups like conservatives, progressives, small business owners, factory workers, or rural communities, depending on the subject. Each agent holds that group's genuine beliefs and self-interest, but has a mandate to be honest and collaborative. They propose, object, revise, and repeat.