Sortition Foundation

The Sortition Foundation is a Cambridge-based not-for-profit social enterprise, informally founded in 2015 and incorporated on 4 March 2017, that provides stratified random selection services for citizens' assemblies and campaigns for their permanent installation in democratic governance. The organization fills the corpus's UK anchor slot for the participant-selection infrastructure layer of deliberative AI and digital governance — the mechanism by which the claim that an assembly genuinely represents the relevant public is constructed and verified. Where organizations such as the Collective Intelligence Project operate on the deliberation design and analysis side of the same practice, the Sortition Foundation operates at the selection infrastructure end: the point at which representativeness is either achieved or not.

Founding and structure

The Sortition Foundation traces its origin to 2015, when a group of sortition advocates met at a London café; nine people formally became directors the following year. Official incorporation followed on 4 March 2017 as company number 10652142 — a private limited company by guarantee without share capital, the legal form used for not-for-profit social enterprises in England. The organization is co-founded and directed by Brett Hennig — an astrophysicist (PhD), former software engineer and taxi driver, and author of The End of Politicians: Time for a Real Democracy — whose trajectory moved from disillusionment with conventional party politics through research into network and deliberative forms of democracy into the practical infrastructure work the Foundation represents.

The Foundation operates with offices in Cambridge UK, Brunswick Australia, Portland USA, and Budapest Hungary. In 2021 its members unanimously adopted a worker-cooperative model through new Articles of Association, giving staff democratic control over organizational decisions — a structural choice that embeds the governance values the Foundation promotes into its own constitution.

Selection methodology and scale

The Sortition Foundation's core capability is stratified random selection: sorting a population of potential participants by demographic variables — age, gender, ethnicity, location, education, and others depending on the assembly's brief — then drawing a random sample whose composition mirrors the commissioning population within specified tolerances. The claim that a citizens' assembly is genuinely representative of a broader public lives entirely in this step. The first major contract came in late 2018 with the Scottish Parliament's Committee Engagement Unit. Growth from this base was rapid: the 2021 Annual Report records 42 citizens' assemblies supported in a single year — matching the total output of the previous two years combined — from approximately 333,000 recruitment letters, which produced 9,000 respondents and over 1,600 participants selected. Since 2017 the Foundation has supported hundreds of deliberative events for local and national governments, the European Commission, NGOs, and universities across the UK, Europe, Australia, and North America.

The methodological foundation received formal academic validation in a 2021 Nature paper — "Fair algorithms for selecting citizens' assemblies" — co-authored by Brett Hennig with Ariel Procaccia (Harvard), Bailey Flanigan, Paul Gölz, and Anupam Gupta (Carnegie Mellon). The paper proves that the LEXIMIN algorithm used in the Foundation's open-source StratifySelect software is the fairest possible selection algorithm: it maximizes the minimum representation probability across demographic groups, resolving a long-standing tension between statistical fairness and panel diversity that earlier heuristic methods could not address. By the time of publication the Foundation had alone deployed LEXIMIN to select over forty panels.

Major assemblies

The most visible assembly the Foundation recruited for is Climate Assembly UK — 110 members assembled across four weekends in spring 2020 at the commission of the UK House of Commons, to produce recommendations on how the UK should reach net-zero greenhouse gas emissions by 2050. The Foundation co-ran the assembly alongside Involve and mySociety. Other significant engagements documented in the 2021 Annual Report include Scotland's Climate Assembly (2020, 105 participants selected from 20,000 invitations); Newham Citizens' Assembly in London, described as the country's first permanent citizens' assembly; the Global Assembly feeding into COP26 (2021); and local government assemblies on subjects spanning assisted dying legislation in Jersey — where the assembly contributed to legislation passing — alcohol pricing, road congestion, and NHS advisory work.

Digital governance and AI dimension

The Scotland National Digital Ethics People's Panel (September 2020) is the corpus's primary documented case of a Sortition Foundation assembly addressing digital and AI governance directly. The panel recruited 30 randomly selected citizens from 6,000 invitations sent to randomly selected Scottish households, to deliberate on "How should Scotland best respond to the digital revolution?" — a framing that encompasses algorithmic decision-making, data governance, and AI-enabled public services as legitimate subjects for public deliberation. The panel's composition was determined by the Foundation's stratified selection process, making it structurally identical to any AI governance assembly: its claim to represent Scotland depended on the same selection infrastructure the Foundation applies to climate, health, and planning assemblies.

More broadly, as deliberative AI governance — the practice of constituting representative public bodies to advise on AI development, deployment, and values — moves from experimental to institutional practice, the Sortition Foundation is the UK practitioner organization with the deepest operational capability to constitute such assemblies. The selection layer is typically invisible in accounts of deliberative AI governance (which focus on the deliberation design and outcome analysis), but it is the load-bearing layer: an assembly that samples from volunteers rather than a stratified random draw of the population is not making the representativeness claim; one that uses a peer-reviewed open-source algorithm and publishes its stratification parameters is. The Foundation's Nature-validated methodology is what that claim requires.

OpenDLP and the infrastructure turn

The Sortition Lab, the Foundation's internal R&D division, launched the Open Democratic Lottery Platform (OpenDLP) as a three-year project beginning January 2025. OpenDLP is an open-source digital platform designed to automate and standardize the full participant selection process — from sending invitations and managing responses through stratification and final lottery draw — making it accessible to commissioning bodies globally without requiring specialist statistical support. The announcement positions OpenDLP as direct infrastructure democratization: since 2017 the Foundation has been one of a small number of organizations with the technical capability to run a genuinely stratified selection at scale; OpenDLP is designed to be the layer that generalizes that capability. The Foundation is also a co-author of the StratifySelect open-source software and the associated panelot.org web interface, allowing anyone to run the LEXIMIN algorithm without installation.

The Foundation's concurrent advocacy campaigns — establishing a permanent Greater Manchester Citizens' Assembly and replacing the House of Lords with a House of Citizens — reflect its institutional thesis that sortition-based assemblies should be permanent democratic infrastructure, not one-off advisory exercises. That thesis directly shapes the AI governance case: a permanent assembly infrastructure capable of advising on digital governance across policy cycles is a structurally different proposition from a single commissioned deliberation, and the Foundation's infrastructure investments are building toward the former.