How The Bias-Variance Tradeoff Shapes Human Strategy Selection

The extent to which biased decision-making is adaptive is one of the longest-debated questions in cognitive science. We investigate whether people adaptively select decision strategies following an optimal bias-variance trade-off: decision strategies that integrate (imperfect) estimates of many uncertain attributes may be unbiased but suffer from high variance, so simpler (biased) decision strategies can be more accurate when uncertainty is high. Using a new multi-attribute choice task, in which we manipulate uncertainty by varying the reliability of the estimates of different attributes across conditions, we provide an empirical demonstration that people indeed adaptively shift their decision strategies in response to environmental uncertainty: participants were more likely to use single-attribute decision-making under high uncertainty and to integrate all attributes under low uncertainty. These findings provide evidence that the bias-variance trade-off is a key principle shaping human strategy selection.