Learning to Act in the Face of Uncertainty

In the study, human volunteers decided whether a cloud of dots presented on a screen was moving to the left or to the right. They held a robotic handle in their right hand and moved the handle towards the target in the direction of their decision (see Figure 1, Figure 2A). The uncertainty of the decision was modulated by changing the coherence level of the dots' motion. When all the dots were moving to the left or to the right (100% coherent motion), the certainty of the decision was high. When only a small proportion of the dots were coherently moving and the other dots were moving in random directions, the situation was uncertain. Participants' hand movement to express the decision was pushed by the robotic handle to deviate from the straight path (see Figure 2B), and under this perturbation, participants were instructed to make a straight movement by resisting the force.

In one of the experiments, participants were divided into two groups: the Certain Decision group and the Uncertain Decision group. The Certain Decision group learned to make a straight movement only after deciding on a high dot coherence level motion (100%). The Uncertain Decision group learned the same action, but only after deciding on a low coherence level motion (3%). Although both groups of participants successfully learned to make a straight movement in their respective decision contexts (certain or uncertain) (see Figure 3A), their performance level dropped when they were asked to perform the same movement after decisions with different uncertainty levels (different motion coherence levels) (see Figure 3B). Participants in the Certain group could resist perturbation at the same level as during their practice after a certain decision, but not after an uncertain decision (see Figure 3B; red line). Similarly, participants in the Uncertain group could perform well after an uncertain decision, but not after a certain decision (see Figure 3B; blue line). This indicates that decision and action are not independent; action is learned in association with the decision that led to the action.

In the other experiment, different types of perturbation (clockwise (CW) direction and counterclockwise (CCW) direction) were each associated with a different decision uncertainty level (see Figure 4B). If the decision uncertainty level does not differentiate the actions that follow the decision, participants should not be able to learn the two perturbations at the same time, since they would interfere with each other. However, if the decision uncertainty level 'tags' the action, participants should be able to learn the two perturbations simultaneously. Participants were indeed able to learn the two perturbations at the same time (see Figure 4A), demonstrating that the actions following certain and uncertain decisions are treated as different things in the brain.

The researchers believe that this finding may open up new avenues for developing novel training methods in sports. By associating skill training with various decision-making situations, athletes can enhance their performance by refining their motor memories in context-specific scenarios.