Search

This checklist is intended to help researchers working at Leiden Universityprepare for carrying out their research data collection in the field, either abroad, or off-site in The Netherlands. Further support for research data management is available at Leiden University from a network of data stewards and data management experts.

While cognitive control allows people to prioritize goal relevant information over habitual or otherwise compelling information, people do not always engage a maximal level of cognitive control. A theorized reason for this is that cognitive control is inherently effortful, and people will avoid exerting effort when it is not sufficiently beneficial (e.g., Shenhav et al., 2013; Kool & Botvinick, 2014; Kool & Botvinick, 2018). This effect has been studied using the demand selection task (DST) which shows that when people are given multiple choices, they will consistently choose less demanding options over more demanding options when reward is equivalent between the two (Kool et al., 2010). Additionally, they will engage more effort when sufficiently compensated (Westbrook & Braver, 2015). This demand avoidance effect has been observed across myriad forms of cognitive control demand (e.g., Cameron et al., 2010; Sayali & Badre, 2019). These demand avoidance effects extend to the cognitive control of attention. Within Stroop tasks, participants avoided choice options where items were mostly incongruent (i.e., the identity of a target picture and distracting word did not match; word CAT printed over an image of a dog) compared to a choice option where items were mostly congruent (i.e., the identity of a target picture and distracting word did match; word DOG printed over an image of a dog) when in a DST (Bustos et al., in prep; see also Schouppe et al., 2014). Indeed, the cognitive control needed to resolve conflict between response options for incongruent trials makes them more demanding than congruent trials. In the current study we are interested in whether switching between control settings (i.e., the degree to which a person processes the target dimension relative to the distractor dimension) is effortful.

It is often asserted that goal-directed planning requires cognitive control. Most evidence for this notion comes from studies showing that cognitive load lowers the propensity to plan towards goals. Here, we aim to demonstrate the opposite. First, using a sequential two-step decision making task, we aim to manipulate the extent by which participants adopt goal-directed planning (i.e., operationalized as model-based reinforcement learning), as compared to a simpler trial-and-error strategy (i.e., model-free reinforcement learning). We hypothesize that maintained reliance on goal-directed control in the decision making task will impair performance on interspersed arithmetic problems. Specifically, in the decision making task we will manipulate the relative difficulty and effectiveness of implementing model-based control by changing the switch rate at which participants (healthy adults) need to update the task’s transition structure that determines the action-outcome contingency. Participants will perform two blocks of the decision making task, with either a high- or low-switch rate (block order randomized). Arithmetic probe trials will be interspersed within the main decision making task, every 10 ± 1 trials. In a pilot study, the high-switch rate block induced a shift towards the model-free strategy, likely due to the increased need to update the transition structure (i.e., increased cognitive demands). Based on the hypothesis that model-based control impairs performance on interspersed arithmetic problem, we expect a decrease in accuracy for the arithmetic problems interspersed in the low-switch rate block, compared to the high-switch block. More generally, we predict that the degree to which people will eventually rely on model-based control in a given block will negatively impact performance on the orthogonal mental calculation problems within the same block.

Cognitive control is vital for regulating behavior. Prior research suggests that participants aim to minimize demands for cognitive control, attaching an effort cost to its exertion. This introduces an intriguing question: How do participants learn these effort costs associated with different lines of actions and apply them during subsequent decision making? Here, we preregister a replication of a pilot study aimed to address this question.

When making a decision, one can use habitual associations or expend effort to use internal models to plan towards goals. Recent work has used a machine learning framework to formalize the distinction between these two systems using model-free and model-based reinforcement learning. This formalization has led to a great deal of progress in modeling how individuals use goal-directed control to plan their actions using cognitive maps. However, it remains largely unknown how people construct these representations of the task’s structure. We have developed a behavioral task aimed at answering this question. In this task, people perform a two-stage decision-making task that can dissociate between model-based and model-free choice. To get an implicit measure of their internal representations of task structure, participants will perform judgements of relatedness between the stimuli that represent the choice options both before and after the two-stage task. Changes in relatedness ratings as a function of learning are used to make inferences about the construction of cognitive maps. In a pilot study, we found the degree to which these relatedness ratings reflected that participants encoded action-outcome relations predicted their reliance on model-based control during the two-stage task. Here, we preregister a set of analyses that aims to replicate these results in a new sample.

This preregistration is for a follow-up version of a previous experiment, which was preregistered at https://osf.io/ujpyb.

Big or small, most decisions incorporate the tradeoff between exploration and exploitation—whether to take advantage of what we know to be good, or to take a chance on something new. Recent research suggests we make these choices via a combination of stochasticity and directedness, and that the directedness involves prioritizing lesser seen options. Here, we preregister a replication study to describe a third form of exploration, one of opportunity detection.

When making decisions involving risk, people tend to be risk-averse when choices are presented in a gain frame and risk-seeking when the same options are in a loss frame (Kahneman & Tversky 1979, Tversky & Kahneman 1981). This observation has been coined the “framing effect” and has led to groundbreaking discoveries in decision-making research. However, it is unknown whether people prefer making decisions when options are in a gain or loss frame and whether this preference is associated with their risk attitude. We have constructed a behavioral paradigm, “The Frame Selection Task,” designed to answer these questions. In the initial phase of the task, participants choose between guaranteed and risky 'gambles' in both frames to measure their risk attitude (following De Martino et al., 2006). In the main phase of the task, participants choose in which frame they want the two gambles to be presented before they choose between them. At the end of the experiment, participants will complete the General Risk Propensity Scale (GRiPS) (Zhang et al. 2018), a questionnaire assessing their risk propensity, so we can determine whether frame preference depends on risk propensity. In a pilot study (n = 46), we found that participants selected the risky gamble more when gambles were presented in a loss frame than in a gain frame in both the initial phase (M = 18%, significantly greater than chance, t(45) = 5.20, p < 0.001) and the main phase (M = 9%, significantly greater than chance, t(45) = 3.00, p < 0.01). We also found that participants preferred making decisions in a gain frame (M = 58%, significantly greater than chance, t(45) = 2.88, p < 0.01) and that people with a stronger preference for the gain frame were more risk-averse (r = 0.33, p < 0.05). We preregister this study to replicate these results in a new sample.