“Evidence, Urgency, and Decision-Making in a Changing World”
Faculty Host: Jordan Taylor
The prevalent model of decision-making suggests that during deliberation, the brain accumulates sensory information to a decision threshold, at which point it commits to a choice. This “drift diffusion model” (DDM) captures a wide range of behavioral data on reaction times and error rates, and suggests that neural activity build-up seen in many brain regions during many decision-making tasks is the neural correlate of evidence accumulation. However, while the DDM is a popular explanation of laboratory experiments, it is not well-suited to the kinds of real-world decisions that the brain evolved to perform. In particular, it does not take into account the redundancy between samples, does not maximize reward rate, and most importantly, is sluggish in responding to changes in the world. As an alternative, I will describe a variation called the “urgency-gating model” (UGM), which suggests that the brain only accumulates novel sensory information and combines the result with a context-dependent signal related to the rising urgency to make a choice. I will show why this model is a better policy for maximizing reward rates in the natural world, and why previous experiments could not effectively distinguish between it and the DDM. I will then describe the results of two experiments that strongly favor the UGM over the DDM. The first is a human experiment, which shows that the brain does not integrate noisy information but behaves like a low-pass filter with a short time constant. The second is a neurophysiological experiment that identifies the neural correlates of evidence, the urgency signal, and the process of decision commitment in the cerebral cortex and basal ganglia of monkeys.