Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Reputation systems are an important part of many markets and online platforms. One central design choice that market designers face pertains to how much feedback to collect, and how to aggregate it. To explore this matter, we vary the design of reputation systems in the lab to evaluate how aggregation choices affect (i) evaluation behavior, (ii) expectations and (iii) rates of cooperation. We find that aggregation matters, even when disaggregated information is also available. Results show that with average evaluations, incentives for reputation building are stronger, evaluations are “more extreme”, expectations are more optimistic, and contributions are about 50% higher.