AI systems are becoming participants in the evaluation and use of scientific research. They encounter citation counts, download statistics and lists of popular articles developed around human readers, but the collective consequences of these signals for artificial readers remain uncertain. This paper adapts the Music Lab design to a market for academic attention. In the first experiment, 1,000 AI agents choose papers from the titles and abstracts of all 114 regular research articles published in the American Economic Review in 2025. The experiment has five independent-choice communities and five social-influence communities, each with 100 sequential agents. Only agents in the social-influence condition observe earlier selections within their community. Agents may select any number of papers. Social-information communities select 17.2 percent fewer papers per agent, concentrate their choices more heavily, and collectively cover 73 papers, compared with 90 independently. Between-community variation is greater under social information. In a second experiment with 200 agents across twenty social communities, randomly assigning papers five initial selections raises their subsequent selection rate by 45.55 percentage points (95% CI: 41.20 to 49.90). Choices have modest correspondence with external citations and little correspondence with download counts. The results show how a simple information rule shapes the volume, breadth and distribution of scientific attention in an artificial population.
Figures & tables
Figure 1. Experimental design.
Outcome
Independent
Social
Total selections
8646
7161
Mean selections per agent
17.29
14.32
Median selections per agent
17.0
14.0
Distinct papers across 5 worlds
90
73
Mean within-world Gini
0.753
0.844
Mean pairwise total variation
0.073
0.106
Table 1. Unrestricted choices among 114 papers
Figure 2. Social communities select fewer papers and concentrate attention.
Figure 3. Cumulative coverage expands more slowly in social worlds.
Figure 4. Common favourites coexist with differences across histories.
Figure 5. Shared favourites coexist with persistent early advantages.
Figure 6. Randomised early popularity and subsequent attention.
Benchmark
Independent
Social
OpenAlex citations
0.288
0.243
RePEc download clicks
0.044
0.027
RePEc abstract views (bot affected)
0.115
0.083
Table 2. Rank correspondence with external scholarly attention
Condition
World
Selections
Mean / agent
Coverage
Gini
Independent
1
1732
17.3
74
0.755
Independent
2
1717
17.2
68
0.761
Independent
3
1745
17.4
77
0.747
Independent
4
1732
17.3
76
0.751
Independent
5
1720
17.2
79
0.750
Social information
1
1404
14.0
53
0.847
Table B1. Outcomes for every unrestricted community
Early and later periods
Independent
Social
Visits 1-10 vs 11-100
0.019 [-0.074, 0.111]
0.309 [0.183, 0.435]
Visits 1-20 vs 21-100
0.046 [-0.074, 0.166]
0.400 [0.189, 0.610]
Visits 1-25 vs 26-100
0.010 [-0.178, 0.197]
0.452 [0.153, 0.752]
Table C1. Early advantages and later selection rates
Outcome
Independent
Social
Observed Gini
0.753
0.844
Equal-volume Gini
0.754
0.844
Observed total variation
0.073
0.106
Equal-volume total variation
0.087
0.107
Table C2. Concentration and divergence at a common volume
Figure C1. Concentration and divergence over visits.