In 2022 we were commissioned by the artificial intelligence company Google DeepMind to help them with some of their research. They were working on an AI tool designed to help groups of people find common ground.
They asked us to recruit and randomly select 300 people (in two batches, one of 100, one of 200) who would help them test the effectiveness of this tool. They wanted a wide range of people from across the UK to participate -- we recruited people in such a way that there was a representative spread across gender, age, ethnicity, area of deprivation and region.
The selected people took part in a number of online sessions across late 2022 and early 2023. Participants did not interact with each other. Instead they were simultaneously asked to give their opinion on a wide range of topics. The AI tool, which the researchers called the "Habermas Machine" then tried to capture the underlying shared perspectives of the participants by writing a "group statement". This group statement was then shared with the participants to see whether they would endorse it -- if not, the Habermas machine could take into account the participants' critique of the statement and have another go.
The research has finally appeared in the prestigious peer-reviewed journal Science. It has been written up so far in the MIT Tech Review, The Guardian, Science news, and Nature news.
We recognise that the effect of AI on the functioning of democracy is a deeply vexed question. In this instance, though, the researchers from DeepMind found reasons for optimism: the final sentence of their conclusion reads "the Habermas Machine offers a promising tool for finding agreement and promoting collective action in an increasingly divided world."