Is the juice worth the squeeze? Learning the marginal value of mental effort over time

Abstract

In keeping with the view that individuals invest cognitive effort in accordance with its relative costs and benefits, reward incentives typically improve performance in tasks that require cognitive effort. At the same time, increasing effort investment may confer larger or smaller performance benefits—i.e., the marginal value of effort—depending on the situation, or context. On this view, we hypothesize that the magnitude of reward-induced effort modulations should depend critically on the marginal value of effort for the given context, and furthermore, the marginal value of effort of a context should be learned over time as a function of direct experience in the context. Using two well-characterized cognitive control tasks and simple computational models, we demonstrate that individuals appear to learn the marginal value of effort for different contexts. In a task-switching paradigm (Experiment 1), we found that participants initially exhibited reward-induced switch cost reductions across contexts—here, task switch rates—but over time learned to only increase effort in contexts with a comparatively larger marginal utility of effort. Likewise, in a Flanker task (Experiment 2), we observed a similar learning effect across contexts defined by the proportion of incongruent trials. Together, these results enrich theories of cost-benefit effort decision-making by highlighting the importance of the (learned) marginal utility of cognitive effort.

Cost–benefit considerations have limited effect on the decision to exert cognitive effort in real-world computer-programming tasks

“…This effect may further indicate how people update their estimation after experiencing the effort associated with function avoidance, i.e. learning the marginal value of mental effort [38], supporting the biased effort-estimation account of avoiding the use of functions.…”

The neural basis of cost-benefit trade-offs in effort investment: a quantitative activation likelihood estimation meta-analysis

“…Given that the region tracking integrated cost-benefits and task demands overlapped, this suggests that a great deal of the high demand tasks used in this meta-analysis were difficult but achievable i.e., high effort was invested for harder tasks. However, this operationalization, which underlies a deal of research on cognitive effort, may not always be satisfied when greater effort does not yield better performance or when participants are given impossible tasks (Otto et al, 2021; Silvestrini et al, 2022).…”

“…Indeed, some previous work has implicitly taken advantage of the marginal value of effort by contrasting performance-contingent to random rewards (Frömer et al, 2021; Shenhav et al, 2013; Späti et al, 2014). On this view, effort should only be invested when increasing effort investment confers larger performance benefits—i.e., the marginal value of effort (Otto et al, 2021). This distinction between effort and demand is reified in established accounts of motivated behavior such as Motivation intensity theory (Brehm & Self, 1989) which posit that the prospect of larger rewards does not unwaveringly improve performance but may depend on the efficacy of effort exertion in improving task performance.…”