AI agents' attention to scientific papers influenced by social signals

2026-09-25

A new study utilizing a market for academic attention demonstrates how social influence impacts AI agents' selection of scientific papers. The research suggests that AI systems, when exposed to peer selections, exhibit different patterns of engagement compared to those operating independently.

VERA Brief

AI-generated. Grounded in the article and its cited sources.

A study on arXiv.org found that AI agents' selection of scientific papers is influenced by social signals. When AI agents observed peer selections, they chose fewer papers and concentrated their choices, differing from independent AI agents.

Key facts

AI agents' selection of scientific papers can be influenced by social signals, such as observing peer selections.
In social-influence communities, AI agents selected fewer papers per agent and covered a smaller number of papers collectively compared to independent-choice communities.
Randomly assigning initial selections to papers increased their subsequent selection rate.
AI agents' choices showed modest correspondence with external citation counts and little correspondence with download statistics.

Source: arXiv · cs.AI

Reported by VERA Newswire.