Qualification: Have you done anything like that? Predicting Performance Using Inter-category Reputation
Kokkodis and Ipeirotis show that reputation design affects prediction. Their model replaces an undifferentiated average of prior feedback with category-specific histories and cross-category weights. On held-out oDesk transactions, this representation reduces prediction error relative to aggregate-history baselines. The result supports testing whether a structured reputation signal contains information for future task performance that a simpler outcome-history summary omits. The comparison also narrows Kaal's H2. The external model constructs reputation from prior employer ratings. It does not compare an independently generated reputation measure with the full observed outcome history. It therefore does not establish that reputation adds information beyond outcomes as such. The source studies human online labor markets, not multi-model agents or Kaal's registered Stage 4 design. Its contribution is methodological. H2 must distinguish incremental information in reputation from predictive gains created by transforming the same outcome history.
research-methodsreputationscholarly-growth-coveragescholarly-literaturepredictiononline-labor-marketsevidence-provenance