One of the most fundamental and striking limitations of hu-man cognitive function is the constraint on the number ofcontrol-dependent processes that can be executed simultane-ously. However, the sources of this capacity constraint re-main largely unexplored. Previous work has attributed the con-straints on control-dependent processing to the sharing of rep-resentations between tasks in neural systems. Here, we exam-ine how shared representations interact with two other factorsin producing constraints on control-dependent processing. Wefirst demonstrate that the detrimental effects of shared repre-sentations on multitasking performance are contingent on theamount of conflict that is induced by the tasks that share rep-resentations. We then examine how the persistence of sharedrepresentations between tasks affects processing interferenceduring serial task execution. Finally, we discuss how this set ofmechanisms can account for various phenomena in neural ar-chitectures, including the psychological refractory period, taskswitch costs, as well as constraints on cognitive control.

It has recently been shown that the cost associated with switching tasks is eliminated following ‘no-go’ trials, in which response selection is not completed, suggesting that the switch cost depends on response selection. However, no-go trials may also affect switch costs by interfering with the effects of task preparation that precede response selection. To test this hypothesis we evaluated switch costs following standard go trials with those following two types of non-response trials: no-go trials, for which a stimulus is presented that indicates no response should be made (Experiment 1); and cue-only trials in which no stimulus is presented following the task cue (Experiment 2). We hypothesized that eliminating no-go stimuli would reveal effects of task preparation on the switch cost in cue-only trials. We found no switch cost following no-go trials (Experiment 1), but a reliable switch cost in cue-only trials (i.e., when no-go stimuli were removed; Experiment 2). We conclude that no-go trials can modulate the switch cost, independent of their effect on response selection, by interfering with task preparation, and that the effects of task preparation on switch cost are more directly assessed by cue-only trials.

Determinantal Point Processes (DPPs) are probabilisticmodels of repulsion, capturing negative dependenciesbetween states. Here, we show that a DPP inrepresentation-space predicts inferential biases towardmutual exclusivity commonly observed in word learning(mutual exclusivity bias) and reasoning (disjunctivesyllogism) tasks. It does so without requiring explicitrule representations, without supervision, and withoutexplicit knowledge transfer. The DPP attempts tomaximize the total ”volume” spanned by the set ofinferred code-vectors. In a representational system inwhich combinatorial codes are constructed by re-usingcomponents, a DPP will naturally favor the combinationof previously un-used components. We suggest thatthis bias toward the selection of volume-maximizingcombinations may exist to promote the efficient retrievalof individuals from memory. In support of this, we showthe same algorithm implements efficient ”hashing”,minimizing collisions between key/value pairs withoutexpanding the required storage space. We suggestthat the mechanisms that promote efficient memorysearch may also underlie cognitive biases in structuredinference.

Cognitive control is critical for accomplishing daily tasks andyet we experience it as effortful or costly. Researchers havebeen increasingly interested in estimating how costly cognitivecontrol is for a given individual, to better understand underly-ing mechanisms and predict motivational impairments outsidethe lab. Here we leverage a computational model of controlallocation to (a) demonstrate a procedure for estimating indi-vidual’s control costs from task performance and (b) highlightthe conditions under which estimated costs will be confoundedwith other motivational variables. We show that costs of cog-nitive control can be reliably estimated under perfect assump-tions about other motivational variables. However, our simu-lation results indicate that poorly calibrated estimates of thoseother variables can lead to potentially drastic misestimations ofsubjects’ control costs, compromising the validity of empiricalobservations. We conclude by discussing the implications ofthese analyses for assessing individual differences in the costsof cognitive control.

Researchers have long been interested in using laboratory mea-sures of cognitive control to predict a person’s cognitive con-trol/self control success outside the lab. We used a computa-tional approach to identify which lab-based performance mea-sures provide the most valid individual difference measuresof one’s ability and/or motivation to exert cognitive control.We simulated performance across an array of cognitive controltasks, and estimated the degree to which different performancemetrics (e.g., congruency effects, conflict adaptation, and de-mand avoidance) could theoretically provide valid estimatesof processes underlying control allocation. By performing di-mension reduction on these performance metrics, we furtherrevealed latent dimensions that can index separate mechanismsof control-demanding behavior. Our results suggest that indi-vidual differences in measures of cognitive control can orig-inate from multiple factors, several of which are unrelated tocapacity for cognitive control. We conclude by discussing im-plications of these analyses for assessing individual differencesin cognitive control phenomena.

Context (such as our location or current goal) informs everyday decisions, both by predicting stimuli and determiningrelevant responses. How do we develop priors that are general enough to apply in various contexts yet specific enough tomaximize reward in a given context? We investigated this using the AX-CPT, a task in which a cue determines which buttonto press for a probe that appears seconds later. We manipulated the frequency of the probe given the cue across participantsand built a diffusion model to estimate how the cue informs participants’ priors for the decision. We found that participants’context-dependent priors were closer to each other and less extreme than those predicted by a model that maximizes rewardrate given the true stimulus frequencies. However, participants’ priors were optimal given their subjective frequency estimates,which showed that they averaged response probabilities across cues when the cues made sufficiently similar predictions.

A striking limitation of human cognition is our inability to ex-ecute some tasks simultaneously. Recent work suggests thatsuch limitations can arise from a fundamental trade-off in net-work architectures that is driven by the sharing of representa-tions between tasks: sharing promotes quicker learning, at theexpense of interference while multitasking. From this perspec-tive, multitasking failures might reflect a preference for learn-ing efficiency over parallel processing capability. We explorethis hypothesis by formulating an ideal Bayesian agent thatmaximizes expected reward by learning either shared or sep-arate representations for a task set. We investigate the agent’sbehavior and show that over a large space of parameters theagent sacrifices long-run optimality (higher multitasking ca-pacity) for short-term reward (faster learning). Furthermore,we construct a general mathematical framework in which ratio-nal choices between learning speed and processing efficiencycan be examined for a variety of different task environments.

Any adaptive organism faces the choice between taking actions with known benefits (exploitation), and sampling new actions to check for other, more valuable opportunities available (exploration). The latter involves information- seeking, a drive so fundamental to learning and long-term reward that it can reasonably be considered, through evolution or development, to have acquired its own value, independent of immediate reward. Similarly, behaviors that fail to yield information may have come to be associated with aversive experiences such as boredom, demotivation, and task disengagement. In accord with these suppositions, we propose that boredom reflects an adaptive signal for managing the exploration-exploitation tradeoff, in the service of optimizing information acquisition and long-term reward. We tested participants in three experiments, manipulating the information content in their immediate task environment, and showed that increased perceptions of boredom arise in environments in which there is little useful information, and that higher boredom correlates with higher exploration. These findings are the first step toward a model formalizing the relationship between exploration, exploitation and boredom.

Daily life often makes us decide between two goals: maximizing immediate rewards (exploitation) and learning about the environment so as to improve our options for future rewards (exploration). An adaptive organism therefore should place value on information independent of immediate reward, and affective states may signal such value (e.g., curiosity vs. boredom: Hill & Perkins, 1985; Eastwood et al. 2012). This tradeoff has been well studied in “bandit” tasks involving choice among a fixed number of options, but is equally pertinent in situations such as foraging, hunting, or job search, where one encounters a series of new options sequentially. Here, we augment the classic serial foraging scenario to more explicitly reward the development of knowledge. We develop a formal model that quantifies the value of information in this setting and how it should impact decision making, paralleling the treatment of reward by the marginal value theorem (MVT) in the foraging literature. We then present the results of an experiment designed to provide an initial test of this model, and discuss the implications of this information-foraging framework on boredom and task disengagement.

A fundamental question in memory research is how different forms of memory interact. Previous research has shown that people rely on working memory (WM) in short-term recognition tasks; a common view is that episodic memory (EM) only influences performance on these tasks when WM maintenance is disrupted. However, retrieval of memories from EM has been widely observed during brief periods of quiescence, raising the possibility that EM retrievals during maintenance-critically, before a response can be prepared-might affect short-term recognition memory performance even in the absence of distraction. We hypothesized that this influence would be mediated by the lingering presence of reactivated EM content in WM. We obtained support for this hypothesis in three experiments, showing that delay-period EM reactivation introduces incidentally-associated information ( context ) into WM, and that these retrieved associations negatively impact subsequent recognition, leading to substitution errors (Experiment 1) and slowing of accurate responses (Experiment 2). fMRI pattern analysis showed that slowing is mediated by the content of EM reinstatement (Experiment 3). These results expose a previously hidden influence of EM on WM, raising new questions about the adaptive nature of their interaction.