Resource‐rational approach to meta‐control problems across the lifespan

(a) Schematic figure of the Flanker task. In this paradigm, participants have to indicate whether the center bee is flying to the left or right. On congruent trials, the flanking bees are flying in the same direction. On incongruent trials, the flanking bees fly in the opposite direction. Before each trial, a reward cue which indicates how many points can be obtained if the correct response is given is shown on the screen (adapted from Devine et al., 2020). (b) Schematic figure of the modified two‐stage decision‐making task. (i) Each trial begins with one of two possible states in which participants must decide between different pairs of spaceships which allows them to transition to either of two second‐stage states (i.e., a red vs. purple planet). The relationship between spaceships and planets is deterministic. Each planet hosts a single alien delivering fluctuating rewards which slowly changes on a trial‐by‐trial basis. (ii) At the start of each trial, participants are presented with a reward multiplier indicating if points on that trial would be multiplied by 1 (low stake trials) or 5 (high stake trials). Reprinted from Kool, Gershman, and Cushman (2017)

Graphical representation of the sweet spot hypothesis. The y‐axis represents the expected reward from engaging in control, which is fixed and stable across age groups. The x‐axis represents the computational costs of applying said strategy, which is exaggerated for older adults and children who have capacity limitations. The z‐axis represents the degree of optimal control one ought to invest given the relative benefits and costs of engaging in cognitive control. The colored curves represent the points at which these three axes intersect for each age group. Shaded areas around each line represent individual variability within each age group. As can be seen from the figure, as computational costs increase, but reward remains stable across age groups, the optimal level of control decreases. Thus, the peak of each curve represents the “sweet spot” in which the relative cost–benefit structure yields optimal cognitive investment. Notice also, however, that these peaks remain lower for children and older adults, reflecting their reduced cognitive capacities

Visual representation of the satisficing model. (a) The model predicts that normative behavior (similar to younger adults) is characterized by updates of predictions to the most likely outcome (mean of the distribution / helicopter position). (b) In contrast, in line with a satisficing model, children and older adults update their predictions until reaching the satisficing threshold, at which time an acceptable position is established and they start perseverating. (c) Example satisficing criteria by age group