Microsimulation to the aid of democracy: the case of Beamm

Liberal democracy faces mounting challenges as polarisation, social media-driven filter bubbles, and AI-enabled disinformation increasingly erode the shared factual reality that any functioning democracy requires as its epistemological foundation. Citizens increasingly inhabit separate factual worlds — and democratic deliberation degenerates into a contest of narratives rather than a debate grounded in commonly agreed consequences.

Tax-benefit policy lies at the heart of these tensions. Fiscal systems are extraordinarily complex and opaque, making fiscal debates particularly vulnerable to ideological distortion in the absence of accessible, shared analytical tools. This keynote argues that microsimulation is uniquely well positioned to help rebuild such a shared factual foundation: by working directly on representative microdata and computing actual rules on actual household heterogeneity, it produces results that are simultaneously accurate, comprehensive, transparent, and communicable — the structural antithesis of the partisan epistemic ecosystem.

We illustrate this democratic potential through the Beamm project (Belgian Arithmetic Microsimulation Model, www.beamm.brussels), a cloud-native, open-access tax-benefit microsimulation platform designed as a public good for democratic deliberation. We discuss in particular the methodological challenges that online and open deployment impose: the need to generate a fully synthetic yet statistically faithful population dataset — combining administrative and survey microdata through statistical matching algorithms and generative AI — that can be hosted in the cloud without compromising the confidentiality of sensitive administrative records. We conclude by arguing that the microsimulation community bears a democratic responsibility, and has the tools, to treat open-access national platforms as a form of international public infrastructure for democratic deliberation.