We propose that self-control failures, and variation across individuals in self-control abilities, are partly due to differences in the speed with which the decision-making circuitry processes basic attributes like taste, versus more abstract attributes such as health. We test these hypotheses by combining a dietary choice task with a novel form of mouse tracking that allows us to pinpoint when different attributes are being integrated into the choice process with millisecond temporal resolution. We find that, on average, taste attributes are processed about 195 ms earlier than health attributes during the choice process. We also find that 13 - 39% of observed individual differences in self-control ability can be explained by differences in the relative speed with which taste and health attributes are processed.

No abstract

Decision research has experienced a shift from simple algebraic theories of choice to an appreciation of mental processes underlying choice. A variety of process-tracing methods has helped researchers test these process explanations. Here, we provide a survey of these methods, including specific examples for subject reports, movement-based measures, peripheral psychophysiology, and neural techniques. We show how these methods can inform phenomena as varied as attention, emotion, strategy use, and understanding neural correlates. Two important future developments are identified: broadening the number of explicit tests of proposed processes through formal modeling and determining standards and best practices for data collection.

The drift diffusion model (DDM) provides a parsimonious explanation of decisions across neurobiological, psychological, and behavioral levels of analysis. Although most DDM implementations assume that only a single value guides decisions, choices often involve multiple attributes that could make separable contributions to choice. Here, we fit incentive-compatible dietary choices to a multi-attribute, time-dependent drift diffusion model (mtDDM), in which taste and health could differentially influence the evidence accumulation process. We found that these attributes shaped both the relative value signal and the latency of evidence accumulation in a manner consistent with participants' idiosyncratic preferences. Moreover, by using a dietary prime, we showed how a healthy choice intervention alters mtDDM parameters that in turn predict prime-dependent choices. Our results reveal that different decision attributes make separable contributions to the strength and timing of evidence accumulation -providing new insights into the construction of interventions to alter the processes of choice. MainSimple choices, like those between food items, have been characterized using sequential integrator models such as the drift (or decision) diffusion model (DDM) 1-4 . In the DDM, choices arise from a process that dynamically integrates evidence for and against each option over time -and a decision is made when the evidence signal reaches the threshold associated with one of the choice options.These models have been enhanced to account for various features of the decision process, allowing them to better explain choices and to generate new insights into cognitive processes. For example, gaze, 5 and pupil dilation, 6 and neural data [7][8][9][10][11] have incorporated the influence of attention and neural signals, resulting in improved predictions. See Ratcliff 1 for a review of advances in the DDM.

The drift diffusion model (DDM) provides a parsimonious explanation of decisions across neurobiological, psychological, and behavioral levels of analysis. Although most DDM implementations assume only one type of information guides decisions, choices often involve multiple attributes that may have differential effects. Here, we fit incentive-compatible dietary choices to a multi-attribute, time-dependent, drift diffusion model (mtDDM), in which taste and health independently influenced the relative value signal that drives the accumulation process in a manner consistent with participants' idiosyncratic preferences. Health information entered the decision process after a longer latency than taste information, diminishing the likelihood of healthy choices. Finally, by using a dietary prime, we showed that variation in mtDDM parameters followed interindividual variation in observed behavior. Our results show that different decision attributes make separable contributions to the timing and strength of evidence accumulationand thus provide new insights into the construction of interventions that may shape the choice process. Figure 1 | Example of the decision process modeling within the multi-attribute, time-dependent drift diffusion model (mtDDM).In this example choice between an indulgent food (tasty but unhealthy) and a disciplined food (healthy but not tasty), a relative value signal (RVS) begins with a value set by a bias parameter (here, set to zero) and evolves only with noise, ε, at every timepoint t as depicted in equation and segment "a". Once the taste attribute's latency is reached (red dotted line), the relative (indulgentdisciplined) taste value, ΔT, begins contributing to the RVS at a rate determined by its drift slope δT (segment "b"). After the health attribute latency is reached (segment "c"), relative health value, ΔH, also begins contributing to the RVS at a rate determined by its drift slope δH. At each time point, Gaussian noise ε is added to the value signal. A decision is said to be reached when the RVS becomes equal to or greater than the boundary for an item. In this example, the taste attribute has a "temporal advantage" (i.e., it begins contributing to the RVS first), and thus the indulgent option boundary is nearly crossed before the health attribute begins contributing toward the eventual disciplined choice. In the example decision process depicted here, a simulated RVS path is displayed for a choice in which the difference in taste and health attributes are ΔT = 1 and ΔH = -7, and mtDDM parameters are set to δT = 0.005 units/ms, δH = 0.0009 units/ms, t*T = 200 ms, t*H = 400 ms, bias = 0, and indulgent option boundary = 1 and disciplined option boundary = -1. Figure adapted from 15 .