Policy abstraction as a predictor of cognitive effort avoidance

Abstract

Consistent evidence has established that people avoid cognitively effortful tasks. However, the features that make a task cognitively effortful are still not well understood. Multiple hypotheses have been proposed regarding which task demands underlie cognitive effort costs, such as time-on-task, error likelihood, and the general engagement of cognitive control. In this study, we test the novel hypothesis that tasks requiring behavior according to higher degrees of policy abstraction are experienced as more effortful. Accordingly, policy abstraction, operationalized as the levels of contextual contingency required by task rules, drives task avoidance over and above the effects of task performance, such as time-on-task or error likelihood. To test this hypothesis, we combined two previously established cognitive control tasks that parametrically manipulated policy abstraction with the demand selection task procedure. The design of these tasks allowed us to test whether people avoided tasks with higher order policy abstraction while controlling for the contribution of factors such as time-on-task and expected error rate (ER). Consistent with our hypothesis, we observed that policy abstraction was the strongest predictor of cognitive effort choices, followed by ER. This was evident across both studies and in a within-subject cross-study analysis. These results establish at least one task feature independent of performance, which is predictive of task avoidance behavior. We interpret these results within an opportunity cost framework for understanding aversive experiences of cognitive effort while performing a task. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Figure 1:

Visual example of levels of policy abstraction and number of options. a) In the first example, the policy abstraction level is 1. Thus, there are four options for each abstraction level. b) In the second example, the policy abstraction level for the task is 2. There are two options at the second abstraction level (red versus white) and two options for each branch at the first level (circle vs. square). This means that in the second example, shape information alone is insufficient to make a response. See the online article for the colored version of this figure.

Figure 2:

Experiment 1 Methods. a) The response task consisted of three conditions in which the number of responses was manipulated. In every trial, a colored box appeared on the screen, and the button was cued to be pressed. b) The feature task consisted of three conditions that manipulated the number of sets of stimulus-response mappings. In each trial, a colored box containing an arrow appeared. The color of the box cued the target direction of the arrow so that the response mapping set was relevant. c) Policy abstraction in Experiment 1. Condition R1 was at the 0^th level of policy abstraction, as no competition was required. The R2 and R4 conditions from the response task and the F1 condition from the feature task are at the 1^st level of policy abstraction, as they each had one level of branching on the presented tree. The F2 and F4 conditions were at the 2^nd level of policy abstraction, as there are two levels of branching in each condition. d) Learning Phase schematic. The task condition symbol was presented first, followed by the participants engaging in 12 practice trials. e) Selection Phase schematic. First, the participants were presented with two task symbols, and they chose the task they would rather execute. The participants then executed the task associated with the symbol. f) Symbols used in Experiment 1. Top row: These symbols were associated with response task conditions, with far left representing the first number of options, middle the second number of options, and far right the third number of options. Second row: These symbols were associated with the feature task conditions. Far left represents the first number of options, middle the second, and far right the third. g) Novel symbols. Left: Novel symbol for the response level, representing an unlearned rule set. Right: Novel symbol for the feature level, representing an unlearned rule set. Hand icons used in this figure are a mirrored version of “Silhouette hand” by Simon Waldherr, which is licensed under CC BY-SA 3.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/3.0/deed.en.

Figure 3:

Results of Experiment 1. a) Reaction times for the Learning and Selection Phases by condition. Conditions are broken into abstraction levels. b) Error rates for the Learning and Selection Phases, by condition. Conditions are broken into abstraction levels. c) Overall selection rates for each condition, with a significant main effect of condition. Conditions are separated by the abstraction level. d) Selected pairwise avoidance rates. Far left: Selection rates for the overall tasks; middle left: control condition (R2 vs. F1); and right: the two abstraction levels. Individuals generally chose response tasks over feature tasks, but were not significantly different from 50% when comparing R2 to F1. The likelihood of choosing an easier option within the abstraction level was not significantly different between the two abstraction levels. The value of 50% is represented by a dashed line. In all the graphs, the dots represent the mean values for each participant. See the online article for the colored version of this figure.

Figure 4:

a) The four conditions, with example stimuli for the baseline, subtask, and return to baseline in the verbal modality. Reprinted from The hierarchical organization of the lateral prefrontal cortex, by Nee, D. and D’Esposito, M., 2016, eLife. b) Schematic of the Learning phase. First, the symbolic representation is presented; next, participants engage in 8–11 trials. c) Schematic of the Selection phase. Participants are presented with two symbolic representations, and they choose which they would rather execute from between the two. Then, participants execute the task associated with that representation. d) The four possible symbolic representations. Top left: The control condition was always associated with a square. Others: The switch, delay, and dual conditions were each randomly assigned to one of these three symbols. See the online article for the colored version of this figure. Task figures in panel a were provided by Derek Nee via personal communication.

Figure 5:

Results for Experiment 2. a) Reaction times for Learning and Selection Phases, by condition. b) Error rates for Learning and Selection Phases, by condition. c) Overall selection rates by condition. In all graphs, shapes represent average values for a participant. See the online article for the colored version of this figure.