Rushed decisions rely on confirmation bias, not less caution

The Speed-Accuracy Trade-off (SAT) is one of the most well-established examples of the adaptability of human decision-making, where faster decisions can come at the expense of accuracy. For simple decisions, the threshold-based account, which says that individuals gather less evidence when responding quickly, is widely accepted. We present a fundamentally different account, arguing that evidence accumulation is more self-excitatory under time pressure, and show that it provides a better explanation of the SAT across six existing datasets (fits to 179 individuals across more than 78,000 trials). We demonstrate that this model better captures key behavioral patterns than the traditional evidence-threshold model, and is more parsimonious than a collapsing-threshold model, which can also account for the data in principle. Finally, we propose that self-excitation emerges from a confirmation bias in attention allocation, such that initially promising evidence is preferentially attended and so drives the decision process when making fast decisions.