THOUGHT EXPERIMENTS IN PUBS

AT NEWSPEAK HOUSE

LIVING WITH TECH

E3: THE WAY WE PARTICIPATE

By Francesca Galli

1 THE LOTTERY

The letter arrives on a Tuesday morning. CAROL, a primary school teacher, opens it expecting a jury summons. Instead it reads: "You have been selected by the National Sortition System to serve on the Citizens' Assembly on Housing Policy. Your three-month term begins on 1 September."

She has not applied for anything. Her name was drawn by a cryptographic lottery. This lottery is mathematically impossible to rig, and it is designed to produce a cross-section of the population that is as “representative” as possible. The process was publicly audited. Nobody chose Carol, the algorithm did.

Her neighbour DAVID, a retired civil servant who has spent thirty years thinking about housing policy, was not selected. He is furious.

Carol's first instinct is to decline. Her second instinct, after reading the briefing pack, is curiosity. She goes.

The Assembly she joins is unlike anything she's encountered. Eighty people — a scaffolder, a retired nurse, a software engineer, a single mother, a farmer — deliberating over six weekends, briefed by experts, asked to produce a policy recommendation. It is, she thinks, what the Athenian kleroterion must have felt like: the ancient stone machine that randomly selected citizens for governance by feeding black and white balls through a tube, in full public view, so everyone could see the process was honest. The Athenians, she learns, considered elections deeply suspicious. They thought that it favoured the rich, the eloquent, the well-connected. To use randomness was instead the democratic option.

The Assembly produces a bold recommendation. The government is not obliged to implement it, but it will get discussed in Parliament.

QUESTIONS TO CONSIDER:

Carol didn't choose to participate, but she was chosen. Is random selection more or less democratic than election?
What would you say democracy is actually for? Representation, optimisation of policy outcomes, or something else entirely?
David knows more about housing policy than almost anyone in the Assembly. Does that matter?
The Athenians considered election undemocratic because it systematically favoured certain people. Do you agree? Does that argument apply to our current system?
Do we need technology (whether digital with cryptography or analog as with the Athenians) in order to enable active participation? Does participation need to be “forced” or at least strongly encouraged?
If you were randomly chosen by a machine to deliberate on an issue you don’t care for, would you still feel obliged to do it?
Would you want to receive Carol's letter? What would make you say yes, or no?

SOURCES & FURTHER READING:

Brian Klaas, Corruptible (2021) — sortition and rotation as practical anti-corruption tools; randomness as a way to hold power accountable rather than just to distribute it

The kleroterion — the Athenian stone sortition machine, a physical and publicly verifiable randomising device; the direct historical precedent for cryptographic sortition

David Spiegelhalter, The Art of Uncertainty (2024) — what "true" randomness means mathematically and why verifiable randomness is both technically hard and politically important

Thought Experiment is written by Francesca Galli, TEiP Group Member, assisted by AI. You can read more of her writing here.

2 DELIBERATION ASSISTANCE

A citizens' assembly has been running for three weekends. The topic is social care funding: who should pay, how much, and for whom. It is hard going. People are talking past each other. Someone starts crying. Two participants have stopped speaking to each other after a disagreement about inheritance tax. The facilitator is exhausted.

On Friday evening, the assembly organisers make an announcement. Two new AI tools are available!

The first tool, the Synthesiser, can process everything said across three weekends — every argument, every concern, every value expressed — and produce a synthesised statement: a draft position that bridges the group's divergent views, identifies where consensus exists, and flags the remaining points of genuine disagreement. It will be ready by Saturday morning. The group can then deliberate from that starting point rather than from scratch.

Some participants are relieved. FATIMA, a care worker who has spoken movingly about her experiences but struggled to translate them into policy language, thinks it could finally help her voice land. GEORGE, a retired barrister, is uneasy. "The disagreements," he says, "are the point. The moment you average them into a consensus statement, you've already made a decision about which disagreements matter. You've just hidden who made it."

The second tool, the Questioner, is different. It does not synthesise or summarise, but instead returns questions. It has identified three perspectives that were voiced strongly but then dropped from the conversation, and three assumptions the group has made that nobody has tested. It can offer these back to the room, and nothing else.

The group has to decide which tool to use, or continue unassisted.

QUESTIONS TO CONSIDER:

Which tool would you choose: the Synthesiser, the Questioner, or neither? Why?
What do you think deliberation is for? Why should we do it in the first place?
Political researchers have identified seven "deliberative muscles": self-reflection, reasoning, dialogue, vulnerability, collaboration, imagination, and facilitation. Which of these can the AI tools strengthen? And what do they substitute for?
Do you have any personal experiences with similar kinds of AI tools?
Fatima's concern is real: some people's voices translate more easily into policy language than others. Could AI help make some perspectives more visible? If so, how?
How do you feel about AI tools potentially becoming standard infrastructure for deliberation? What are the potential benefits, or risks?
What would you want participants to be able to do — and to be — at the end of a deliberative process that they couldn't do or be at the start?

SOURCES & FURTHER READING:

Claudia Chwalisz, Sammy McKinney, Jorin Theuns & Eugene Yi, Deliberative Muscles and AI (DemocracyNext, 2026)

Shutaro Aoyama, People Don't Know What They Want (2026)

Tessler et al., AI can help humans find common ground in democratic deliberation (Science, 2024)

Thought Experiment is written by Francesca Galli, TEiP Group Member, assisted by AI. You can read more of her writing here.

3 DIGITAL IDENTITY

MARGARET is seventy-one. For forty years she used the same post office to collect her pension, renew her bus pass, and access emergency support. The woman behind the counter knew her name. When Margaret's husband died, she quietly mentioned a bereavement payment that Margaret didn't know she was entitled to.

A few years ago the post office closed. Everything moved online.

The new government platform is, in many ways, genuinely good. When Margaret registered her bus pass, it automatically flagged she was likely eligible for a winter fuel supplement she had never claimed. She told her daughter it was the first time a government service had felt like it was on her side.

Three months later, her pension didn't arrive. Her account had been suspended by an automated fraud-detection system, which had flagged her login behaviour as anomalous (she logs in slowly, takes a long time on each page). The decision hadn’t been reviewed by any human being. The appeal process required a biometric photograph uploaded via a smartphone app. Margaret does not own a smartphone. The helpline had no option to speak to a person.

JAMES is thirty-four, works in tech, and finds the same platform impressive. Last week he renewed his driving licence, updated his address, and checked his eligibility for a home insulation grant in one twenty-minute session.

Over lunch, a colleague mentions that the data infrastructure underneath the platform (the identity systems, the registers, the APIs) is operated by a private company under a seven-year contract, with a clause about sharing data with "approved analytics partners." James, who understands data architecture better than most people using the platform, realises he has no idea what "approved" means or who decides. He had assumed someone had worked this out. He is not sure who that “someone” is supposed to be.

He thinks about this for a moment, then closes the tab and goes back to work.

QUESTIONS TO CONSIDER:

The post office worker noticed Margaret was struggling before she said anything. Is this something a public service should offer? Could a digital system be designed accordingly?
Margaret's account was suspended by an algorithm working exactly as designed. Who is responsible for what happened to her? The designers, the government, the private contractor, the politicians who commissioned it?
Margaret "owned" her credentials — her pension entitlement, her bus pass, her heating support eligibility. But when her account was suspended, she couldn't access any of them. Does she actually “own” her digital identity?
Think about what digital identity means to you. In what ways is it a reflection of who you are and what you do, and in what ways does it fall short?
James understands data architecture, but he hadn't thought about who owns the infrastructure or what the data-sharing clause means. Is his “informed consent” meaningful?
The infrastructure of citizenship (the systems that determine who gets a pension, who gets flagged as a fraud risk, etc.) is operated by a private company under contract. Does that matter?
What would genuine public control of the citizenship tech infrastructure look like? How might we all participate in that? Should we?

SOURCES & FURTHER READING:

Richard Pope, Platformland (2024) — on the design of digital public services

Palantir's Federated Data Platform contract with NHS England — £330m contract for a US defence contractor to manage NHS patient data; the live parallel to James's story

The UK Post Office Horizon scandal — subpostmasters criminalised by data from a faulty private IT system treated as authoritative and unchallengeable

Jamie Susskind, Future Politics (2018) — code as law; the privatisation and automation of force; terms of service as private legislation

Thought Experiment is written by Francesca Galli, TEiP Group Member, assisted by AI. You can read more of her writing here.