From Polarization to Policy
Using Human Digital Twins to discover policies people can agree on.
The challenge
How can a policymaker design effective policies when public debate starts from deeply polarized positions? On issues such as ecological transition, migration or inequality, traditional research can accurately measure disagreement. But it is much harder to identify new policy formulations capable of creating broad consensus across different demographic and political groups.
The HDTwin approach
HDTwin created a virtual panel of 1,000 Human Digital Twins, designed from a representative sample of the Italian population provided by IPSOS. Starting from highly divisive statements, the Digital Twins participated in an iterative AI-assisted deliberation: they voted, expressed different perspectives, and evaluated new statements generated during the discussion. Instead of simply measuring existing opinions, HDTwin explored where convergence could emerge. The process progressively identified principles and formulations capable of bridging initially opposing positions.
From consensus to policy
The areas of convergence discovered through the virtual deliberation were then translated into concrete policy proposals. In the case of ecological transition, one of the resulting proposals focused on promoting Renewable Energy Communities, combining environmental goals, economic benefits for citizens, and easier access to renewable energy. When tested on the real-world IPSOS sample, the proposal achieved 94.1% agreement among Italian respondents. Across the broader experiment, AI-assisted consensus statements increased human agreement by more than 37 percentage points compared with the original divisive statements, while policies inspired by the deliberation increased support by more than 25 percentage points.
A new workflow for policy design
The approach creates a new workflow for policy design: simulate → discover convergence → design policy → validate with real people.
A synthetic deliberation laboratory
HDTwin does not replace citizens or policymakers. It provides a synthetic deliberation laboratory in which thousands of interactions and alternative formulations can be explored before moving to real-world validation. 1,000 Human Digital Twins. Highly divisive starting positions. A real-world policy reaching 94.1% agreement. HDTwin helps policymakers move from measuring polarization to discovering common ground.
Published research
The methodology and results are described in the research paper E Pluribus Unum: AI-assisted Consensus Building by Leonardo Becchetti, Giovanni Cerase, Enrico Fagnoni and Stefano Quintarelli. The paper is freely available for download on SSRN.
Presented on prestigious international stages
IFDaD 2025 — International Forum on Digital and Democracy, AI and Democracy Session at the World AI Cannes Festival, Palais des Festivals et des Congrès, Cannes — February 14, 2025. Festival Nazionale dell’Economia Civile 2025, Democrazia Aumentata. AI, Intelligenza relazionale e nuove architetture del consenso, Palazzo Vecchio, Florence — October 4, 2025. The Florence presentation brought together IPSOS President Nando Pagnoncelli and Copernicani President Stefano Quintarelli to discuss how AI-assisted deliberation can contribute to more open and constructive consensus-building.
Watch the presentation from the Festival Nazionale dell’Economia Civile
Download the research paper
30-minute online conversation; invite policy, research, and stakeholder leads and bring a draft policy question; next step is a consensus-mapping walkthrough.