Discerning Mouse Trajectory Features With the Drift Diffusion Model

Abstract

Mouse tracking, a new action-based measure of behavior, has advanced theories of decision making with the notion that cognitive and social decision making is fundamentally dynamic. Implicit in this theory is that people's decision strategies, such as discounting delayed rewards, are stable over task design and that mouse trajectory features correspond to specific segments of decision making. By applying the hierarchical drift diffusion model and the Bayesian delay discounting model, we tested these assumptions. Specifically, we investigated the extent to which the "mouse-tracking" design of decision-making tasks (delay discounting task, DDT and stop-signal task, SST) deviate from the standard "keypress" design of decision making tasks. We found remarkable agreement in delay discounting rates (intertemporal impatience) obtained in the keypress and mouse-tracking versions of DDT (ρ = 0.90) even though these tasks were given about 1 week apart. Rates of evidence accumulation converged well in the two versions (DDT, ρ = .86; SST, ρ = .55). Omission/commission error in SST showed high agreement (ρ = .42, ρ = .53). Mouse-motion features such as maximum velocity and AUC (area under the curve) correlated well with nondecision time (ρ = −.42) and boundary separation (ρ = .44)-the amount of information needed to accumulate prior to making a response. These results indicate that the response time (RT) and motion-based decision tasks converge well at a fundamental level, and that mouse-tracking features such as AUC and maximum velocity do indicate the degree of decision conflict and impulsivity.

Hand movement trajectories illustrate the mechanism underlying Kurt Lewin’s distinction between approach–approach and avoidance–avoidance motivational conflicts.

“…We computed conflict-strength measures for each of the trajectories: (a) Maximum Deviation (MD; Boschet et al, 2022; Cummins & De Houwer, 2021; Dale & Duran, 2011; Freeman, 2014; Freeman & Ambady, 2009; Freeman et al, 2008; Garcia-Guerrero et al, 2023; Maldonado et al, 2019; McKinstry et al, 2008; Stillman et al, 2017)—the largest perpendicular deviation of the actual trajectory from an ideal straight line between the start and end points, and (b) Area under the Curve (AUC; Boschet et al, 2022; Cummins & De Houwer, 2021; Dale et al, 2007; Freeman & Ambady, 2010; Garcia-Guerrero et al, 2023; Leontyev & Yamauchi, 2021; Spivey et al, 2005; Stillman et al, 2017, 2020)—the geometric area between the actual trajectory and an ideal line. We further computed two measures corresponding to the mechanistic properties of the force-fields theory: (c) x-flips (to measure oscillatory behavior; Dale & Duran, 2011; Dale et al, 2008; Freeman, 2014; Freeman & Ambady, 2010; Freeman & Johnson, 2016)—the number of reversals of direction along the x -axis, and (d) Returns to Point of Balance (RPB; to quantify the extent to which participants were stuck in the middle between the options in a given trial)—the number of returns to the line where the x coordinate equals 0 (Koop & Johnson, 2013).…”

“Wait Short Time” as a Simple and Resource-rational Boost: Theoretical and Empirical Approaches in the Contexts of Speed-accuracy Trade-off

“…Previous studies indicated that the area under the curve (AUC) of mouse cursor movements can be used as an index of the extent of cognitive conflicts [30][31][32] . The AUC is defined as the amount of area between the actual trajectory and a straight trajectory from trial onset to response termination (Figure 4; e.g., 33 ).…”

“One-second wait”: A simple and effective intervention to boost judgment accuracy

“…Maximum velocity time. Maximum velocity time is defined as the time at which the maximum velocity of a mouse movement is achieved in a trial (e.g., 25,39,40). People are expected to move a mouse quickly when they judge which alternatives to choose in their minds.…”