How a free flow of information can amplify incorrect ideas

The idea that information should flow freely is deeply embedded in the design of social media. The assumption is that the more information is produced and shared, the better. However, simulations by a team of scientists including University of Groningen Professor of Artificial Intelligence Davide Grossi show that such an unrestricted flow of information can amplify incorrect ideas amongst like-minded people. The study was published in Proceedings of the National Academy of Sciences on 1 April.

Information sharing is always beneficial. That is the central premise of digital communication platforms. If users want to communicate with one another, no amount of shared information is too much. But is this central premise correct? A team of social scientists and computer scientists used digital agents that shared unlimited information with perfect honesty to study how this affected the development of correct and incorrect ideas.

Reduced accuracy of group beliefs

‘Strikingly, our model suggests that even when agents are honest, cooperative, and have perfect information-processing abilities, allowing them to exchange unlimited information can make the group’s collective beliefs less accurate,' explains Davide Grossi. The simulations worked with binary information (true or false). Furthermore, the agents were homophilous, meaning that they tended to interact with like-minded agents.

‘In the model, the world is in a binary state,’ says Grossi. ‘For example: it rains, or it does not rain. Furthermore, the agents believe that it rains with a certain degree of probability.’ When exchanging information, agents who believe it is raining are more likely to communicate with one another and end up believing more strongly that it is indeed raining. Grossi: ‘This effect would be less severe if agents were somewhat restricted in the number of observations they could exchange.’ This shows that unconstrained information exchange can reduce the accuracy of group beliefs, particularly in socially homogenous settings.

Democratic principles

Furthermore, in such settings, the group with an erroneous belief is driven further away from the truth because they keep exchanging observations that point towards an erroneous understanding of the world. The group with the correct belief, on the other hand, moves more robustly towards recognizing the correct state of the world. Grossi: ‘This means there is a form of polarization occurring due to the combined working of homophily and unrestricted circulation of observations, because groups with similar beliefs tend to amplify their own opinions.’

These findings in digital agents offer a warning for digital communication platforms, according to Grossi and his colleagues: ‘Because these negative effects emerge even in agents that process information perfectly and cooperatively, there is a plausible risk of even more severe effects in less ideal real-world contexts.’

Grossi concludes that rigorous scientific methods are more needed than ever to understand the complex social phenomena brought about by social media, and support the design of platforms that better adhere to democratic principles. ‘Understanding how digital communication platforms affect collective beliefs and behaviour is essential for the development of digital tools that better serve our societies.’

Reference: Jonas Steina, Shannon Cruz, Davide Grossic, and Martina Testori: Free information disrupts even Bayesian crowds. Proceedings of the National Academy of Sciences, 1 April.