When AI meets democracy: earning public trust by design

Themes: Design choices as a form of democratic accountability | Evidence and legitimacy | Transparency and accountability

Public consultations are vital to democracy, but the government struggles to properly analyse what citizens say at scale. 

This talk explores how the government’s Incubator for AI built Consult, its AI tool for analysing consultation responses, and what they learned when they took democratic accountability seriously from day one. 

Explore topics from how they measured quality honestly, to how they earned - not assumed - public trust, to a transparency case that means the process remains accountable if things go wrong.

This is a practical story about what it really takes to deploy AI responsibly in a democratic process.

Participant takeaways:

• how to use human variance benchmarking - measuring the natural inconsistency between human analysts - to create a defensible, realistic standard for AI performance in qualitative research
• the methodology behind the AI Social Readiness Advisory Label: you’ll leave with a practical framework for partnering with independent researchers (like Nesta) to move public trust from a "feeling" to an externally-verified metric
• a step-by-step look at how to build a transparency case for AI in the public sector: learn how to document and publish your methodology so that when things go wrong, the service - and the democracy it supports - remains accountable