Decision making by urgency gating: theory and experimental support

Abstract

It is often suggested that decisions are made when accumulated sensory information reaches a fixed accuracy criterion. This is supported by many studies showing a gradual build up of neural activity to a threshold. However, the proposal that this build up is caused by sensory accumulation is challenged by findings that decisions are based on information from a time window much shorter than the build-up process. Here, we propose that in natural conditions where the environment can suddenly change, the policy that maximizes reward rate is to estimate evidence by accumulating only novel information and then compare the result to a decreasing accuracy criterion. We suggest that the brain approximates this policy by multiplying an estimate of sensory evidence with a motor-related urgency signal and that the latter is primarily responsible for neural activity build up. We support this hypothesis using human behavioral data from a modified random-dot motion task in which motion coherence changes during each trial.

The urgency-gating model can explain the effects of early evidence

“…1d). We have shown support for this latter prediction in recent experiments, first using a task in which subjects made decisions based on discrete events (Cisek et al, 2009), and later using a variation of the randomdot task in which the motion coherence was changing during each trial (Thura et al, 2012). We have also shown that while monkeys make decisions about changing sensory information, neural activity in motor regions tracks evidence quickly (with old evidence leaking out within 200 ms), and combines it with a growing urgency signal .…”

“…It is wellknown that RT distributions in decision-making tasks are very Including the low-pass filter of the UGM, however, does in fact render it responsive to changes in early evidence, and, overall, matches the behavioral data very well broad, and this can be explained either with large intra-trial noise -as in the DDM -or with large inter-trial variability -as in the LATER model (Carpenter & Williams, 1995) and the UGM (Cisek, Puskas, & El-Murr, 2009;Thura et al, 2012). Importantly, variability in the urgency signal also provides an alternative explanation for the shift in mean RT distributions reported by Winkel et al…”

“…In contrast, a low-pass filter with a short time constant can respond more quickly to changes in sensory evidence while remaining robust to noise. For this reason, we and others have suggested that sensory evidence is not integrated continuously but rather low-pass filtered and combined with an independent signal related to the urgency for making a choice -and that together these bring neural activity to a decision threshold (Cisek, Puskas, & El-Murr, 2009;Ditterich, 2006;Thura et al, 2012). We have called this the Burgency-gating model( UGM).…”

“…In the original UGM paper (Cisek et al 2009), we compared four versions of evidence integrator models and two versions of the urgency-gating model, one without (Bmodel 5^) and one with a low-pass filter (Bmodel 6"). The former was meant solely as an illustrative step, and we emphasized the importance of low-pass filtering in the original 2009 paper, as well as in subsequent publications (Thura et al, 2012;Thura, Cos, Trung & Cisek, 2014). Unfortunately, it seems this caused some confusion.…”

A Dynamic Approach to Recognition Memory

“…In two-choice situations with stationary evidence distributions, it is optimal to accumulate information until one of two fixed boundaries is reached (Wald, 1945(Wald, , 1947Stone, 1960;Edwards, 1965;Bogacz et al, 2006). However, in situations where there is a stochastic response deadline (Frazier & Yu, 2008) or one is limited by total time and not number of trials in a task, it is optimal to allow response boundaries to collapse, reflecting the diminishing returns on collecting additional evidence (Thura, Beauregard--Racine, Fradet, & Cisek, 2012;Drugowitsch, Moreno-Bote, Churchland, Shadlen, & Pouget, 2012). The memory evidence posited by our model can be seen as imposing a kind of stochastic deadline, since no new information accrues once the probe is saturated with features.…”