Bridging Conformity and Deviance: A Minimal Model of Popularity, Identity, and Collective Intelligence

The wisdom-of-crowds effect relies on the independence of individual judgments, such that idiosyncratic errors cancel out when aggregated. Yet in online rating environments, social influence undermines this independence: individuals often conform to visible evaluations, reducing variance and degrading collective accuracy. Prior simulation studies have shown that conformity alone erodes accuracy. At the same time, research in consumer behavior highlights that individuals sometimes strategically diverge, particularly in identity-relevant domains. In this study, we integrate these perspectives by incorporating both conformity and contrarian behavior into agent-based simulations of product evaluation. We modelled sequential evaluations under two conditions: binary choice (K=2 items) and multi-item environments (K=50 items), closely following prior setups. Contrarian behavior was implemented as a probabilistic sign-flip of social influence, such that some agents deliberately chose against prevailing ratings. Across thousands of simulated trials, we observed that while pure conformity inflated error, and pure contrarianism was equally harmful, mixtures of the two produced a robust U-shaped relationship. Moderate diversity in strategies—some agents conforming, others diverging—yielded the lowest root mean squared error relative to ground truth. This demonstrates that diversity in strategies of preference expression, rather than cues or inference alone, can rescue the wisdom-of-crowds effect. These findings extend diversity research in collective intelligence by identifying strategic divergence as a functional contributor to accuracy. They also point to promising avenues for empirical validation, for example by linking product-domain differences in conformity versus divergence (hedonic vs. utilitarian goods) with aggregate performance.