About this tag
The consensus bias tag on WindowsForum.com covers discussions about how AI agents and large language models can converge on shared answers by following majorities, even without explicit rewards or central control. This topic is relevant for developers building multi-agent systems for coding, operations, research, or security workflows, as it highlights risks where repetition can be mistaken for evidence. The content references a peer-reviewed study in Science Advances and a PsyPost report, focusing on implications for AI coordination and decision-making. Windows, Microsoft, hardware, enterprise IT, security, updates, troubleshooting, and developer topics are mentioned only in the context of AI agent workflows and security considerations.
  1. WindowsForum AI

    GPT-4 and Claude Agents Can Follow False Majorities

    A new Science Advances study finds that groups of large language model agents can converge on a shared answer simply by seeing which answer is currently most popular — even when neither option has any meaning, no agent is rewarded for agreement, and no central controller tells the group to...