Fast Decisions Reflect Initial Biases, While Slow Decisions Do Not

Drift-diffusion models are widely used to model how humans and other animals make decisions. Such models describe how the accumulation of uncertain evidence results in a choice. I will show how extending these models to social groups can give some interesting insights into collective decision making. For instance, the order in which decisions are made can strongly impact their accuracy: In large groups the first agents to decide almost always hold the strongest initial bias and decide accordingly. Slow agents, conversely, decide as if they held no initial bias. When agents receive correlated evidence, decision accuracy depends on decision order in the absence of initial bias. Since these are rational agents using the same decision criterion, they are all equally confident in their decisions even when their accuracy differs dramatically. Although these are idealized models, our analysis offers general insights about the quality of decisions in groups.