Subjects adjust criterion on errors in perceptual decision tasks.

Abstract

The optimal strategy in detection theory is to partition the decision axis at a criterion C, labeling all events that score above C "Signal", and all those that fall below "Noise." The optimal position of C, C*, depends on signal probability and payoffs. If observers place their criterion at some place other than C*, they suffer a loss in the Expected Value (EV) of payoffs over the course of many decisions. We provide an explicit equation for the degree of loss, where it is shown that the falloff in value will be steep in contexts of good discrimination and will be a flatter gradient in contexts of poor discrimination. It is these gradients of loss in EV that, in theory, drive C toward C*, strongly when discrimination is good, weakly when discrimination is poor. When signal probabilities or distributions variances are unequal, the basins of attraction are asymmetric, so that dynamic adjustments in C will be asymmetric, and thus, as we show, will leave it biased. We address our analysis to acquisition speed, response variability, discrimination reversal and other aspects of discriminated performance. In the final section, we develop an error correction model that predicts empirically observed deviations from C* that are inconsistent with the standard model, but follow from the proposed model given knowledge of d'. (PsycINFO Database Record

Adaptive Choice Biases in Mice and Humans

“…In our 2AFC tasks, participants obtained reinforcement by perceiving and choosing an S D with predictive value. This contingency established a direct link between stimulus discriminability and choice biases ( Killeen et al, 2018 ). In contrast, stimulus discriminability did not affect the production of choice biases when we removed the predictive value of the S D stimulus in the side-choice phase of the [2AFC→ 2AUC] tasks.…”

“…This observation could derive from the simple fact that stimulus strength directly enables the adjustments in the strategies used by decision-makers in 2AFC tasks. At the extreme, with low or zero discriminability, responding exclusively to one side or the other, or even alternating between sides, are equally valid strategies ( Killeen et al, 2018 ). However, an additional consideration is that choice biases could also be influenced by other factors that are not directly linked to the explicit properties of the sensory stimuli.…”

An Automated Water Task to Test Visual Discrimination Performance, Adaptive Strategies and Stereotyped Choices in Freely Moving Mice

“…Another important property of stereotyped behavior is that it is strongly dependent on the choice history as this could offer better adaptive strategies to the inherent imbalances in the natural world (Busse et al, 2011 ; Trevino, 2014 ; Akrami et al, 2018 ). The emergence of such biases could be driven by the lack of discriminative information (Killeen et al, 2018 ), but they could also reflect internal imbalanced processes of individuals (Trevino, 2014 ).…”

“…Stereotyped choices constitute an example of an adaptive behavioral strategy that occurs regularly in psychophysical experiments (Prusky et al, 2000 ; Busse et al, 2011 ; Trevino et al, 2013 ; Trevino, 2014 ). Stereotypical behavior is a term used to describe a wide variety of invariant behaviors that could maximize utilities (Killeen et al, 2018 ) but that could also derive from inherent properties of individuals (Trevino, 2014 ). In mice, stereotyped choices are strongly dependent on reward and sensory histories, they are consistent across animals and determine their learning trajectories (Trevino, 2014 ; Akrami et al, 2018 ).…”

Untitled

“…41 Although work with ordinal categorical judgements does suggest a superiority of TSD as early as 42 1978 (Kornbrot, 1978), since then, much theoretical work has concentrated on development of 43 either the signal detection (Killeen, Taylor, & Treviño, 2018) or the choice framework (Bohil, et 44 al., 2015). In spite of their similarities, the generic mechanisms that lead to logit and probit 45 distributions are different.…”