Tallying heuristic

The tallying heuristic is a fast-and-frugal heuristic that counts the number of cues favoring each alternative and chooses the one with the higher count. It ignores cue validity, cue weight, and cue interactions. It is the simplest compensatory member of the heuristic family: unlike the take-the-best heuristic, which stops at the first discriminating cue, tallying integrates all cues equally. It is fast because it does not require ranking cues by validity; it is frugal because it does not require estimating cue weights. It is the decision strategy of someone who counts pros and cons without asking which pro or which con matters more.

Despite its simplicity, tallying performs remarkably well across a wide range of environments. In simulation studies, it has matched or exceeded multiple regression in predictive accuracy, particularly in environments where cues are roughly equally valid and where the goal is to predict a criterion rather than to explain it. The reason is that the additional complexity of weighted models often overfits the data: the weights are tuned to the sample, not to the underlying structure. Tallying, by ignoring weights, avoids this overfitting. It is a robust strategy — not optimal in any single environment, but adequate across many. The ecological rationality of tallying lies in its robustness, not in its specificity.

The tallying heuristic has implications for democratic decision-making, jury trials, and committee design. If the best group decision strategy is often a simple, unweighted vote, then the elaborate weighting schemes of representative democracy may be not merely inefficient but structurally misguided. The heuristic also challenges the logic of expert panels: if simple tallying outperforms weighted expert judgment, then the appropriate role of expertise is not to assign weights but to identify cues. The cognitive engineering of collective decision-making may require less aggregation and more cue identification. The less-is-more effect applies to groups as well as individuals: the group that integrates all opinions equally may outperform the group that weights opinions by expertise.