AI Agents Concentrate Research Attention via Social Influence
September 22, 2026
An experiment with 1,000 AI agents shows that social-information communities select 17.2% fewer papers and concentrate choices more heavily than independent agents. This suggests that citation and popularity signals may create feedback loops when AI agents participate in scientific evaluation.
HOW THIS AFFECTS YOU
●
researcherYou should account for agent-driven popularity biases in automated literature reviews.
●
policyThis highlights potential risks in how algorithmic reputation affects scientific progress.