Dynamic choices are most accurate in small groups

Abstract

According to the classic results of Galton and Condorcet, as well as in modern decision-making models, accuracy in groups increases with group size. However, these studies do not consider the naturally occurring situation in which individuals dynamically re-evaluate their decision with a possible change of opinion. The dynamics of re-evaluation in groups are very different to individual re-evaluation because individuals influence the group and the group influences the individual. We find that individual accuracy in a group is higher when individuals re-evaluate because all members have more access to social information, while in single decisions, those deciding first have less. This improvement is smaller in large groups as in this case errors can cascade across the members of the group before re-evaluation can correct them. The net result is a maximal accuracy at a small group size. We also analyzed the case in which individuals are influenced only by a small number of the other individuals. In this case, cascading errors affect the interacting subgroups but are very unlikely to reach the whole group. This results in a local optimum at a small group size but also an optimum at a very large size. We thus

Modular structure within groups causes information loss but can improve decision accuracy

“…In natural environments, animals typically do not make judgments completely independently of each other, as assumed in our analysis so far, and in models such as the Condorcet jury theorem. Instead, the opinions of animals are often correlated with each other to some degree, and in such scenarios, small group sizes have often been shown to maximize collective accuracy [16,18]. This is because increasing group size initially increases collective accuracy owing to the benefit of opinion aggregation (the ‘wisdom of crowds’), but at larger group sizes, the correlated cue increasingly dominates the collective decision, decreasing accuracy.…”

A Simple Computational Theory of General Collective Intelligence

“…Alignment on utility functions may be more difficult to achieve in diverse groups, but then, perhaps if an ultimate consensus utility is achieved it could be better for more people outside the group. Group size has also been a topic of interest (Kao & Couzin, ; Vicente‐Page, P′erez‐Escudero, & Polavieja, ). It is plausible that alignment is easier to achieve in small groups due purely to statistical effects of distributions of beliefs and preferences.…”

Probabilistic social learning improves the public’s judgments of news veracity

“…As a mechanism, it has been found that probabilistic response scales enrich social learning by allowing individuals to explicitly signal uncertainty in their judgments [20][21][22][23][24][25][26][27]. By contrast, exchanging coarse-grained binary judgments can prevent individuals from signaling their uncertainty-for example, when voting 'yes' or 'no' on whether an event will occur, different people may vote 'no' with either high or low confidence; the binary signal alone does not distinguish among them [28][29][30]. Moreover, probabilistic response scales allow individuals to signal even minor belief adjustments during the communication process that can help steer the group toward a more accurate collective judgment [20][21][22][23][24][25][26][27].…”