China News Service, Beijing, July 5th (Reporter Sun Zifa) Springer Nature's professional academic journal "Nature-Human Behavior" recently published a sociology paper. Through a proof-of-concept study, researchers found that artificial intelligence ( AI) algorithms may be able to propose new mechanisms for allocating resources among the population, and their resource allocation decisions are more popular.

　　Philosophers, economists and political scientists have been divided over the years over how the benefits of human cooperation should be distributed, the paper says.

　　Corresponding author of the paper, Christopher Summerfield of British artificial intelligence firm Deepmind, and colleagues collaborated to train an artificial intelligence system to design a new mechanism for the redistribution of public goods, named "democratic AI" ( Democratic AI).

The researchers first asked thousands of volunteers to participate in an investment game in groups of four.

In the game, each person will get a different amount of money, and they need to decide whether to keep the money for their own use, or share it for the benefit of the entire group, so that the shared funds can be returned to themselves with the principal and interest.

　　The authors of the paper trained the AI ​​system to find a policy that reallocated funds to individuals, whose popularity was determined by the votes of human players when deciding which policy to choose to play again.

The AI ​​system managed to spot a policy that got more votes than the baseline policy, such as redistributing funds evenly to everyone, or returning it in proportion to individual contributions.

When the researchers asked other human volunteers to act as reassignment decision makers, none of their strategies were as popular as the AI's.

　　The authors of the paper note that while the study focused on a four-player special-edition game of public goods, future research could continue to expand the concept of "democratic AI" and investigate whether it is effective for larger groups and more complex game scenarios. .

(Finish)