How the Brain Decides When to Work and When to Rest: Dissociation of Implicit-Reactive from Explicit-Predictive Computational Processes
Figure 4
Introspection of cost evidence.
A–B: Behavioral tasks. The illustrated screenshots were successively presented every trial. A: the Cost Rating Task was developed to assess introspection of resource exhaustion. On each trial, participants were asked to squeeze the hand grip up to the target level (horizontal bar), which corresponded to varying difficulty level (40% to 60% of maximal force), as long as the thermometer was displayed, which could last for varying durations (4 to 7 seconds). After each effort, participants rated their degree of exhaustion using a visual horizontal analog scale. The last screen of each trial indicated the payoff cumulated over preceding trials. B: the Effort Allocation Task was exploited in a previous paper [10]. When the thermometer image was displayed, participants could squeeze a handgrip to win as much money as possible. Subjects were provided with online feedback on force level and cumulative payoff. The payoff was only increased when force level was above the target bar, at a constant rate proportional to the monetary incentive. The incentive (10, 20 or 50 cents) and the difficulty (i.e. the force required to reach the target bar: 70, 80 or 90% of maximal force) were crossed over trials. The last screen indicated the money won over all preceding trials. C–D: Computational modeling: Bayesian model comparison. For each participant, we estimated eight models generating cost evidence from difficulty and duration. Cost evidence was then used to fit the subjective ratings of exhaustion (C) or the decisions to stop effort exertion (D). Models were linear combinations of different possible regressors (main effects, interaction and non-linear effects), as indicated in the bottom chart (tick: included, cross: not included). E–F: Fit of additive and multiplicative models. Data are subjective ratings of exhaustion (E) and probability of stopping effort exertion (F), shown in the duration by difficulty space explored in the Cost Rating and Effort Allocation tasks, respectively. The color code indicates average observed data (left diagram) or predicted data from the additive and multiplicative models that provided the best fit with median parameter values (middle and right diagrams). Note the main difference between additive and multiplicative models is the curvature of iso-value lines (in white), which reflects the interaction between duration and difficulty.