Cognitive And Sensorimotor Interactions In Human Decision-Making Within A Virtual Environment

There is a long tradition of studying economic decision making, where humans often fail to maximise expected gain. More recently, attention has been directed towards decision making in mathematically equivalent sensorimotor tasks, where humans often approach maximum expected gain. But numerous everyday tasks have ‘cognitive’ and ‘sensorimotor’ costs. This raises a fundamental, but hitherto neglected research question about the factors that influence decision making when an economic choice has sensorimotor risks. We created a ‘game’ in virtual reality where participants needed to hit targets in order to win points. The game required participants to choose between two targets where one was easier to hit (closer and on permanent display) and the alternative was a harder-to-hit ‘risky’ target worth more points (further away and programmed to time-out). The time allowed to hit the ‘risky’ target was the median of the individual’s baseline trials. Participants deceased their movement time during the baseline trials so the risky targets were more likely to be hit than not regardless of their distance (this resulted in the risky targets having a higher expected gain with respect to the extrinsic reward). In Experiment 1, we found participants (n = 40) were motivated by the reward (so frequently selected the higher value target). Nevertheless, the behaviour was also influenced by the sensorimotor costs, such that participants were more likely to choose the safe option (despite this decreasing expected gain) when the high reward target (worth twice as many points) was further. We found gender differences whereby women were less likely to reach for the high reward target when it was further away. Subsequently, the same selection frequencies were found in two separate groups (both n = 40) despite the high reward target having three and five times more points than the safe option, suggesting that a sensorimotor cost threshold acts as an upper bound on the selection choice process. In Experiment 2, we added motor noise whilst keeping the expected gain constant and found that this manipulation did not affect decision making (i.e. we found same selection frequencies as in Experiment 1). In Experiment 3, we added perceptual noise and again found that this did not affect the decision making. Experiments 2 and 3 suggest that adults are well tuned to the costs of their sensorimotor actions. The data from all 200 participants showed a bias to: (i) select a risky target after a safe trial; (ii) select a risky target after a high reward target was hit (compared to when it was missed). These behavioural phenomena are well captured by a partially observable Markov decision process (pom-dp), and a pom-dp model was able to capture the behaviour by integrating extrinsic rewards and sensorimotor costs in a choice selection process. The pom-dp predicted that participants should increasingly select the risky target across multiple sessions, with the result that males and females should converge on similar selection rates across the different target distances. Experiment 4 tested this prediction with participants repeating the task across multiple sessions over three days. This resulted in an increased probability of the high reward target being selected, and by the end of the sessions the gender differences were not observed. The first four experiments always contained a known ‘safe’ target so Experiment 5 introduced a selection task where the choices needed to be made in a more dynamic fashion and there was not always an obvious ‘safe’ target. Experiment 5 confirmed that participants rapidly combine extrinsic rewards and sensorimotor costs in order to choose between targets on a trial-by-trial basis. Experiment 6 investigated decision making in younger children and showed that the combination of extrinsic rewards and sensorimotor costs occurs in even 7-8 year old children (though there was greater evidence of sub-optimal selections occurring on some trials when the age of the group was younger).