Redistributive policy

DeepMind’s AI develops popular policy for distributing public money

DeepMind researchers have trained an AI system to find a popular policy for distributing public funds in an online game – but they also warn against ‘AI government’


July 4, 2022

Could artificial intelligence make better funding decisions than senators?

Walter Bibikow/Getty Images

A “democratic” AI system learned to develop the most popular policy of redistributing public money between people playing an online game.

“Many of the problems that humans face are not simply technological, but require us to coordinate in society and in our economies for the greater good,” says Raphael Koster at DeepMind, an artificial intelligence company based in the United Kingdom. “For AI to help, it must learn human values ​​directly.”

The DeepMind team trained their artificial intelligence to learn from over 4,000 people as well as computer simulations in a four-player online business game. In the game, players start with different amounts of money and must decide how much to contribute to help develop a pool of public funds, eventually receiving a share of the pot in return. Players also voted on their preferred policies for distributing public money.

The policy developed by the AI ​​after this formation has generally attempted to reduce wealth disparities between players by redistributing public money according to how much of their starting pot each player has contributed. It also discouraged free-riders by giving players almost nothing back unless they had contributed about half of their starting funds.

This AI-designed policy won more votes from human actors than an “egalitarian” approach of redistributing funds equally regardless of how much each person contributes, or a “libertarian” approach. » consisting of distributing the funds according to the proportion of the contribution of each person. the public pot.

“One thing we found surprising was that the AI ​​learned politics that reflects a mix of viewpoints from all political walks of life,” says Christopher Summerfield at DeepMind.

When there was the greatest inequality between players at the start, a “liberal egalitarian” policy – which redistributed money according to the proportion of starting funds each player contributed, but did not discourage free-riders – proved to be as popular as the AI ​​proposal, receiving over 50% of the votes in a head-to-head contest.

DeepMind researchers warn that their work does not represent a recipe for “AI government”. They say they have no plans to create AI-powered tools for policymaking.

Maybe that’s just as well, because the AI ​​proposal isn’t necessarily unique from what some people have already suggested, says Annette Zimmerman at the University of York, UK. Zimmermann also cautioned against focusing on a narrow idea of ​​democracy as a system of “preference satisfaction” to find the most popular policies.

“Democracy is not just about winning, it’s about implementing the policy you prefer, it’s about creating processes in which citizens can meet and deliberate with each other on an equal footing. equality,” says Zimmermann.

DeepMind researchers worry about an AI-powered “tyranny of the majority” situation in which the needs of people from minority groups are ignored. But that’s not a big worry for political scientists, says Mathias Risse at Harvard University. He says modern democracies face a bigger problem: “the many” are being disenfranchised by the small minority of the economic elite and abandoning the political process altogether.

Still, Risse says the DeepMind research is “fascinating” in how it delivered a version of the liberal politics of egalitarianism. “Since I’m in the liberal-egalitarian camp anyway, I find this a rather satisfying result,” he says.

Journal reference: Nature Human behavior, DOI: 10.1038/s41562-022-01383-x

Learn more about these topics: