Nesta’s Centre for Collective Intelligence, in collaboration with the UK government’s Incubator for Artificial Intelligence, has trialled an approach to involve the public in assessing AI tools for public services.

As AI is increasingly used in UK public services, there’s growing concern about whether the public finds these tools acceptable. We built and tested a new “public-in-the-loop” process to assess the social acceptability of different AI tools used in UK public services. Using deliberative polling through our Zeitgeist platform, 140 people evaluate an AI tool risks and benefits, informed by expert analysis. Their feedback is summarised into a simple Social Readiness Advisory Label to guide decision-makers.

Results from the first AI tool to be assessed by the public through this process, found solid but conditional support: 60% of participants said they were positive about the tool being used in public services. All recommendations for the developers of the tool (DSIT’s Incubator for AI) have been accepted by them. This includes testing alternative AI models to ChatGPT-4o, measuring the environmental impact, and developing strong mitigations against model manipulation. The deliberation also showed statistically significant increases in participants’ comfort with AI being used in public services (+12%); increase in trust in the public sector to deploy AI responsibly (+11%); and belief that public sector AI use will benefit society (+16%).

“The quality of the recommendations demonstrated the quality of the conversation which Nesta facilitated with the public.” Incubator for AI, DSIT