Analysis of hand kinematics reveals inter-individual differences in intertemporal decision dynamics

Abstract

During intertemporal decisions, the preference for smaller, sooner reward over larger-delayed rewards (temporal discounting, TD) exhibits substantial inter-subject variability; however, it is currently unclear what are the mechanisms underlying this apparently idiosyncratic behavior. To answer this question, here we recorded and analyzed mouse movement kinematics during intertemporal choices in a large sample of participants (N = 86). Results revealed a specific pattern of decision dynamics associated with the selection of “immediate” versus “delayed” response alternatives, which well discriminated between a “discounter” versus a “farsighted” behavior—thus representing a reliable behavioral marker of TD preferences. By fitting the Drift Diffusion Model to the data, we showed that differences between discounter and farsighted subjects could be explained in terms of different model parameterizations, corresponding to the use of different choice mechanisms in the two groups. While farsighted subjects were biased toward the “delayed” option, discounter subjects were not correspondingly biased toward the “immediate” option. Rather, as shown by the dynamics of evidence accumulation over time, their behavior was characterized by high choice uncertainty.

Appendix

This Appendix summarizes the methods used in the Embodied Choice model to calculate the action focus; for further details, see Lepora and Pezzulo (2015). Mathematically, we represent the action focus as a point (x (z), y (z)) that is a function of the accumulated information z(t). At each instance of time, the action is a move (Δx, Δy) from the present location toward the action focus

$$(\Delta x(t),\Delta y(t)) = v\Delta t\frac{{(x(z),y(z)_{\text{focus}} - x(t),y(t))}}{{\left| {(x(z),y(z)_{\text{focus}} - x(t),y(t))} \right|}}$$

which for simplicity we assume is at constant speed v taken at discrete time steps of Δt.

In the two-target forced choice task considered here, we consider a preparatory move toward an action focus between the two targets, to approach the most likely target prior to accumulating sufficient information to make a decision. Mathematically, we define the focus as collinear with the two targets with distance from each in the proportion ∣b + z∣: ∣b − z∣ for −b ≤ z ≤ b

$$(x(z),y(z))_{\text{focus}} = \left\{ {\begin{array}{ll} {(x_{1} ,y_{1} ),} &\quad {{z}({t}) \ge {b}} \\ {\frac{{\left| {b - z} \right|}}{2b}(x_{1} ,y_{1} ) + \frac{{\left| {b + z} \right|}}{{{{2}}{b}}}(x_{2} ,y_{2} )} &\quad { - b \le z(t) \le b} \\ {(x_{2} ,y_{2} )} &\quad {{z}({t}) \ge - {b}} \\ \end{array} } \right.$$

This range is bounded such that the focus is coincident with a target if the decision bound is passed. Hence, the decision bound no longer defines decision termination directly, but rather the choice follows indirectly from action completion (upon reaching a target).