KITA AI: Have agents representing distinct social groups deliberate and treat disagreement as an outcome

Public decision-making tools generally provide a single and consensual opinion. This system allows for debate between several models anchored in demographic personas and assumes non-convergence.

Public policies dealing with social and environmental emergencies must take into account multiple, often contradictory perspectives. However, computer-based decision-support devices, even when they incorporate systems of varying values, continue to make a single result, based on a consensus search. KITA AI, presented on September 28, 2026, takes the problem from the other end.

The system is modular: several agents based on language models deliberate on public policy scenarios, each anchored in a distinct demographic persona and within a specific conceptual framework. Its objective is not limited to designating a preferred proposal: it automatically reveals who is affected by the scenario, and provides policymakers with the quantitative reasoning and indicators that support each position.

The most notable design choice is lack of agreement. KITA AI treats non-convergence as a fully explicable result, not as a failure to reduce. The idea is to allow public officials to grasp the compromises involved and the human consequences of the policies discussed, rather than to receive a recommendation whose arbitrations would remain invisible.