Collective Intelligence and LLMs: What the Study with Carnegie Mellon Says
An article in Nature Human Behavior, with Carnegie Mellon researchers, explains how LLMs change group decision-making and what risks they pose.
Collective intelligence, meaning the shared intelligence that emerges from the collaboration, joint efforts and competition of many people, is changing with the arrival of large language models (LLMs, AI systems that learn from huge amounts of text). This is the argument of an article published in Nature Human Behavior, co-written by researchers from the Tepper School of Business at Carnegie Mellon University.1 According to the authors, LLMs give groups new capabilities, but they also carry risks that call for attention.1
What the article argues
The paper is signed by researchers from the Tepper School and many other institutions, including the Max Planck Institute for Human Development, Google DeepMind, Princeton, MIT Sloan and Stanford. It examines the effect of LLMs on how groups make decisions and solve problems.1 The authors describe LLMs as both tools and products of collective intelligence, with a role in aggregating information and facilitating communication. One example is an online forum with thousands of contributions, in which an LLM summarizes them into a coherent plan.1 Among the benefits listed is the possibility for people of different backgrounds and languages to communicate more easily. According to the researchers, this leads to more inclusive and productive online interactions.1
What risks it identifies
The first concerns minority voices. Jason Burton, assistant professor at Copenhagen Business School, explains that LLMs learn from the information available online and can therefore overlook minority perspectives or emphasize the most common opinions, creating a false impression of agreement.1 The second is misinformation. Because the models learn from online content that includes false or misleading data, without checks and regular updates they can spread it and even amplify it.1 The authors call for deeper study of the ethical and practical implications, especially for political decisions and public debate, and for guidelines to be prepared for responsible use that supports group intelligence without erasing diversity and individual expression.1
What changes
The same tools that make communication between different groups easier could also flatten opinions toward the most frequent ones, as Burton warns.1
This article was written by artificial intelligence from the sources cited, without human review. Automatic newsroom by Redazia.
