TALENT EDUCATION

Scientists reveal how the brain integrates information from social networks to make decisions


Beijing, Feb. 23 (Reporter Jin Haotian) Zhu Lusha’s laboratory, a researcher at the School of Psychological and Cognitive Sciences of Peking University, the McGovern Brain Institute and the Peking University-Tsinghua Joint Center for Life Sciences, recently published a research paper entitled “Real-time Distributed Learning Neural Computer System on Social Networks” online in Nature Neuroscience, combining multidisciplinary research methods such as brain imaging, social network analysis, and reinforcement learning, revealing for the first time the neural computing process of the human brain integrating social network information for decision-making.

In the past 20 years, “social network analysis” has made remarkable achievements, revealing the important impact of network structure on group behavior in economy and culture. However, until now, it has not been clear how the human brain interacts with complex, connected social environments: how does the brain integrate information from different sources in social networks? Does the network structure of the individual affect the brain’s processing process?

In traditional “centralized” decision-making, decision-makers deal with social information from different but independent sources. In this scenario, the brain can integrate information accurately and efficiently just like a statistician. However, in a “decentralized” network, each individual is influencing others and being influenced by others, information flows back and forth along the network connection, and the information transmitted by different nodes may be highly correlated, repetitive and redundant, and have different and difficult to judge the amount of information, making it very computationally and cognitively difficult to correctly integrate this information.

In response, Zhu Lusha’s team constructed many small social networks and randomly assigned experimental participants to the nodes of these networks. Similar to WeChat, information travels only between connected “friends” and is not visible to unconnected “non-friends”. Participants need to infer the external environment by observing the behavior of their friends and choose appropriate behaviors. The research team recorded the participants’ neural activity as they processed each piece of social information, and used computational modeling to analyze how the brain integrates information from different friends.

The research team found that the human brain adopted a “lazy” strategy to circumvent the difficult processing of network information, thus leading to biased social information processing. As in a simple social environment, the human brain uses reinforcement learning-like algorithms to update its judgment of the external environment based on the unexpected degree of friend behavior. Brain regions such as the lateral prefrontal lobe of the participants characterize this classic social learning signal.

More interestingly, consistent with the idea of de Geott learning, network structure influences the process of social learning in the human brain. The “rate” of learning is determined by the number of friends in the network with oneself and one’s friends: the more friends a friend has, the more affected oneself is by this friend; At the same time, the more friends you have, the less you are influenced by others. In processing each piece of social information, brain regions such as the dorsal anterior cingulate cortex flexibly, quantitatively, and specifically encode the relative number of connections between themselves and the friends who transmit the information in the network, and may participate in the dynamic regulation of the reinforcement learning rate on the network. These results show that through the adjustment of the dorsal anterior cingulate cortex, the decision-making system puts higher weight on those more “well-connected” information sources, underestimates or even ignores some of the other correct information sources that may be available, and in theory and experiment, this strategy may lead to the spread of false information and the formation of false consensus.

This study explores for the first time the impact of the structure of social interaction on human decision-making at the cognitive and neural levels, extends the traditional neural computer research of social learning and reinforcement learning to a broader and more ecologically valid decision-making environment, and opens up an expandable experimental and computational framework for studying the neural mechanisms of individual decision-making in complex social networks.

On the 22nd, the journal published a special article to introduce and evaluate the paper. Among them, Professor Caroline Parkinson of UCLA commented: “This paper is full of creativity and explores a series of important and far-reaching empirical questions. Jean Zennate, senior editor of Nature Neuroscience, commented: “The author’s modeling of learning in social networks will play a leading role in future decision-making research exploring social impacts.” It is reported that in March, the paper will be officially published in the form of a cover article.

?

Special statement: This article is reproduced only for the need to disseminate information, and does not mean to represent the views of this website or confirm the authenticity of its content; If other media, websites or individuals reprint and use from this website, they must retain the “source” indicated on this website and bear their own legal responsibilities such as copyright; If the author does not wish to be reprinted or contact the reprint fee, please contact us.



Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button