Planning Complexity Registers as a Cost in Metacontrol

Simplifying social learning

Social learning is complex, but people often seem to navigate social environments with ease. This ability creates a puzzle for traditional accounts of reinforcement learning (RL) that assume people negotiate a tradeoff between easy-but-simple behavior (model-free learning) and complex-but-difficult behavior (e.g., model-based learning). We offer a theoretical framework for resolving this puzzle: although social environments are complex, people have social expertise that helps them behave flexibly with low cognitive cost. Specifically, by using familiar concepts instead of focusing on novel details, people can turn hard learning problems into simpler ones. This ability highlights social learning as a prototype for studying cognitive simplicity in the face of environmental complexity and identifies a role for conceptual knowledge in everyday reward learning.

Variable, sometimes absent, but never negative: Applying multilevel models of variability to the backward crosstalk effect to find theoretical constraints

When performing two tasks at the same time, the congruency of the second task's features influences performance in the first task. This is called the backward crosstalk effect (BCE), a phenomenon that influences both theories of binding and of dual-task capacity limitations. The question of whether the BCE is found in all participants at all times is relevant for understanding the basis of the effect. For example, if the BCE is based on strategic choices, it can be variable, but if it is automatic and involuntary, it should never vary in whether it is present or not. Variability in observed BCE sizes was already documented and discussed when the group average effect was first reported (Hommel, 1998). Yet the theories discussed at the time did not motivate a more direct analysis of this variability, nor did the readily available statistical tools permit it. Some statistical approaches recently applied in cognitive psychology allow such a variability-focused analysis and some more recent theoretical debates would benefit from this as well. We assessed the variability of the BCE as well as a BCE-like free-choice congruency effect by applying a Bayesian multilevel modeling approach to the data from a dual-tasking experiment. Trials consisted of a two- and a four-response task. We manipulated which task was presented first and whether the response to the four-choice task was free or forced choice. RT data were best predicted by a model in which the BCE is zero in part of the population and drawn from a normal distribution truncated at zero (and thus always positive) in the rest of the population. Choice congruency bias data were best predicted by a model assuming this effect to be drawn from a normal distribution truncated at zero (but, in contrast to the RT data, without the subset of the population where it is zero). The BCE is not an inflexible and universal phenomenon that is directly linked to an inherent structural trait of human cognition. We discuss theoretical constraints implied by these results with a focus on what we can infer about the traits of the factors that influence BCE size. We suggest that future research might add further major constraints by using multi-session experiments to distinguish between-person and within-person variability. Our results show that the BCE is variable. The next step is understanding along which axes it is variable and why it varies.

Wagers for work: Decomposing the costs of cognitive effort

Some aspects of cognition are more taxing than others. Accordingly, many people will avoid cognitively demanding tasks in favor of simpler alternatives. Which components of these tasks are costly, and how much, remains unknown. Here, we use a novel task design in which subjects request wages for completing cognitive tasks and a computational modeling procedure that decomposes their wages into the costs driving them. Using working memory as a test case, our approach revealed that gating new information into memory and protecting against interference are costly. Critically, other factors, like memory load, appeared less costly. Other key factors which may drive effort costs, such as error avoidance, had minimal influence on wage requests. Our approach is sensitive to individual differences, and could be used in psychiatric populations to understand the true underlying nature of apparent cognitive deficits.

Neurophysiological and Autonomic Correlates of Metacognitive Control of and Resistance to Distractors in Ecological Setting: A Pilot Study

In organisational contexts, professionals are required to decide dynamically and prioritise unexpected external inputs deriving from multiple sources. In the present study, we applied a multimethodological neuroscientific approach to investigate the ability to resist and control ecological distractors during decision-making and to explore whether a specific behavioural, neurophysiological (i.e., delta, theta, alpha and beta EEG band), or autonomic (i.e., heart rate-HR, and skin conductance response-SCR) pattern is correlated with specific personality profiles, collected with the 10-item Big Five Inventory. Twenty-four participants performed a novel Resistance to Ecological Distractors (RED) task aimed at exploring the ability to resist and control distractors and the level of coherence and awareness of behaviour (metacognition ability), while neurophysiological and autonomic measures were collected. The behavioural results highlighted that effectiveness in performance did not require self-control and metacognition behaviour and that being proficient in metacognition can have an impact on performance. Moreover, it was shown that the ability to resist ecological distractors is related to a specific autonomic profile (HR and SCR decrease) and that the neurophysiological and autonomic activations during task execution correlate with specific personality profiles. The agreeableness profile was negatively correlated with the EEG theta band and positively with the EEG beta band, the conscientiousness profile was negatively correlated with the EEG alpha band, and the extroversion profile was positively correlated with the EEG beta band. Taken together, these findings describe and disentangle the hidden relationship that lies beneath individuals' decision to inhibit or activate intentionally a specific behaviour, such as responding, or not, to an external stimulus, in ecological conditions.

Model-free decision making resists improved instructions and is enhanced by stimulus-response associations

Human behaviour may be thought of as supported by two different computational-learning mechanisms, model-free and model-based respectively. In model-free strategies, stimulus-response associations are strengthened when actions are followed by a reward and weakened otherwise. In model-based learning, previous to selecting an action, the current values of the different possible actions are computed based on a detailed model of the environment. Previous research with the two-stage task suggests that participants' behaviour usually shows a mixture of both strategies. But, interestingly, a recent study by da Silva and Hare (2020) found that participants primarily deploy model-based behaviour when they are given detailed instructions about the structure of the task. In the present study, we reproduce this essential experiment. Our results confirm that improved instructions give rise to a stronger model-based component. Crucially, we also found a significant effect of reward that became stronger under conditions that favoured the development of strong stimulus-response associations. This suggests that the effect of reward, often taken as indicator of a model-free component, is related to stimulus-response learning.

Social Concepts Simplify Complex Reinforcement Learning

Humans often generalize rewarding experiences across abstract social roles. Theories of reward learning suggest that people generalize through model-based learning, but such learning is cognitively costly. Why do people seem to generalize across social roles with ease? Humans are social experts who easily recognize social roles that reflect familiar semantic concepts (e.g., "helper" or "teacher"). People may associate these roles with model-free reward (e.g., learning that helpers are rewarding), allowing them to generalize easily (e.g., interacting with novel individuals identified as helpers). In four online experiments with U.S. adults (N = 577), we found evidence that social concepts ease complex learning (people generalize more and at faster speed) and that people attach reward directly to abstract roles (they generalize even when roles are unrelated to task structure). These results demonstrate how familiar concepts allow complex behavior to emerge from simple strategies, highlighting social interaction as a prototype for studying cognitive ease in the face of environmental complexity.

When a thinker does not want to think: Adding meta-control into the working model

De Neys proposes an elegant solution to several theoretical problems of the dual-process theories but underspecifies the role of motivation in initiating, intensifying, and ceasing deliberation. Therefore, I suggest including a meta-cognitive control component in the working model that can moderate deliberation, for instance by affecting the deliberation threshold.

Further advancing fast-and-slow theorizing

The 34 commentaries on the target article span a broad range of interesting issues. I have organized my reply around five major themes that seemed to emerge: Remarks about the generalizability of the empirical findings, links with other models, necessary extensions, the utility of dual-process models, and more specific points. This allows me to clarify possible misconceptions and identify avenues for further advancement.

Neurocognitive basis of model-based decision making and its metacontrol in childhood

Human behavior is supported by both goal-directed (model-based) and habitual (model-free) decision-making, each differing in its flexibility, accuracy, and computational cost. The arbitration between habitual and goal-directed systems is thought to be regulated by a process known as metacontrol. However, how these systems emerge and develop remains poorly understood. Recently, we found that while children between 5 and 11 years displayed robust signatures of model-based decision-making, which increased during this developmental period, there were substantial individual differences in the display of metacontrol. Here, we inspect the neurocognitive basis of model-based decision-making and metacontrol in childhood and focus this investigation on executive functions, fluid reasoning, and brain structure. A total of 69 participants between the ages of 6-13 completed a two-step decision-making task and an extensive behavioral test battery. A subset of 44 participants also completed a structural magnetic resonance imaging scan. We find that individual differences in metacontrol are specifically associated with performance on an inhibition task and individual differences in thickness of dorsolateral prefrontal, temporal, and superior-parietal cortices. These brain regions likely reflect the involvement of cognitive processes crucial to metacontrol, such as cognitive control and contextual processing.

Rethinking model-based and model-free influences on mental effort and striatal prediction errors

A standard assumption in neuroscience is that low-effort model-free learning is automatic and continuously used, whereas more complex model-based strategies are only used when the rewards they generate are worth the additional effort. We present evidence refuting this assumption. First, we demonstrate flaws in previous reports of combined model-free and model-based reward prediction errors in the ventral striatum that probably led to spurious results. More appropriate analyses yield no evidence of model-free prediction errors in this region. Second, we find that task instructions generating more correct model-based behaviour reduce rather than increase mental effort. This is inconsistent with cost-benefit arbitration between model-based and model-free strategies. Together, our data indicate that model-free learning may not be automatic. Instead, humans can reduce mental effort by using a model-based strategy alone rather than arbitrating between multiple strategies. Our results call for re-evaluation of the assumptions in influential theories of learning and decision-making.

Reinforcement learning with associative or discriminative generalization across states and actions: fMRI at 3 T and 7 T

The model-free algorithms of "reinforcement learning" (RL) have gained clout across disciplines, but so too have model-based alternatives. The present study emphasizes other dimensions of this model space in consideration of associative or discriminative generalization across states and actions. This "generalized reinforcement learning" (GRL) model, a frugal extension of RL, parsimoniously retains the single reward-prediction error (RPE), but the scope of learning goes beyond the experienced state and action. Instead, the generalized RPE is efficiently relayed for bidirectional counterfactual updating of value estimates for other representations. Aided by structural information but as an implicit rather than explicit cognitive map, GRL provided the most precise account of human behavior and individual differences in a reversal-learning task with hierarchical structure that encouraged inverse generalization across both states and actions. Reflecting inference that could be true, false (i.e., overgeneralization), or absent (i.e., undergeneralization), state generalization distinguished those who learned well more so than action generalization. With high-resolution high-field fMRI targeting the dopaminergic midbrain, the GRL model's RPE signals (alongside value and decision signals) were localized within not only the striatum but also the substantia nigra and the ventral tegmental area, including specific effects of generalization that also extend to the hippocampus. Factoring in generalization as a multidimensional process in value-based learning, these findings shed light on complexities that, while challenging classic RL, can still be resolved within the bounds of its core computations.

Adaptive search space pruning in complex strategic problems

People have limited computational resources, yet they make complex strategic decisions over enormous spaces of possibilities. How do people efficiently search spaces with combinatorially branching paths? Here, we study players’ search strategies for a winning move in a “k-in-a-row” game. We find that players use scoring strategies to prune the search space and augment this pruning by a “shutter” heuristic that focuses the search on the paths emanating from their previous move. This strong pruning has its costs—both computational simulations and behavioral data indicate that the shutter size is correlated with players’ blindness to their opponent’s winning moves. However, simulations of the search while varying the shutter size, complexity levels, noise levels, branching factor, and computational limitations indicate that despite its costs, a narrow shutter strategy is the dominant strategy for most of the parameter space. Finally, we show that in the presence of computational limitations, the shutter heuristic enhances the performance of deep learning networks in these end-game scenarios. Together, our findings suggest a novel adaptive heuristic that benefits search in a vast space of possibilities of a strategic game.

Search problems usually have a common trade-off between accuracy and computational resources; Finding the best solution usually requires an exhaustive search, while limiting computations usually decreases the quality of solutions. Yet, humans provide high-quality solutions for complex problems despite having limited computational resources. How do they do that? Here, we analyze people’s behavior in a strategic game of “k-in-a-row” that has an enormous space of possibilities. We find that people strongly prune the search space by using scoring strategies to evaluate each possibility and augment this pruning with a shutter heuristic that limits their search to the possible winning paths from their last move. Similar to other adaptive heuristics, the shutter heuristic provides a strong reduction in the computations the searcher needs to carry out while maintaining on par accuracy rates. Finally, this adaptive heuristic generalizes to the performance of deep learning networks when playing with limited computational resources.

Suboptimal human inference can invert the bias-variance trade-off for decisions with asymmetric evidence

Solutions to challenging inference problems are often subject to a fundamental trade-off between: 1) bias (being systematically wrong) that is minimized with complex inference strategies, and 2) variance (being oversensitive to uncertain observations) that is minimized with simple inference strategies. However, this trade-off is based on the assumption that the strategies being considered are optimal for their given complexity and thus has unclear relevance to forms of inference based on suboptimal strategies. We examined inference problems applied to rare, asymmetrically available evidence, which a large population of human subjects solved using a diverse set of strategies that varied in form and complexity. In general, subjects using more complex strategies tended to have lower bias and variance, but with a dependence on the form of strategy that reflected an inversion of the classic bias-variance trade-off: subjects who used more complex, but imperfect, Bayesian-like strategies tended to have lower variance but higher bias because of incorrect tuning to latent task features, whereas subjects who used simpler heuristic strategies tended to have higher variance because they operated more directly on the observed samples but lower, near-normative bias. Our results help define new principles that govern individual differences in behavior that depends on rare-event inference and, more generally, about the information-processing trade-offs that can be sensitive to not just the complexity, but also the optimality, of the inference process.

People use diverse strategies to make inferences about the world around them, often based on limited evidence. Such inference strategies may be simple but prone to systematic errors or more complex and accurate, but such trends need not always be the rule. We modeled and measured how human participants made rare-event decisions in a preregistered, online study. The participants tended to use suboptimal decision strategies that reflected an inversion of the classic bias-variance trade-off: some used complex, nearly normative strategies with mistuned evidence weights that corresponded to relatively high choice biases but lower choice variance, whereas others used simpler heuristic strategies that corresponded to lower biases but higher variance. These relationships illustrate structure in suboptimality that can be used to identify systematic sources of human errors.

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

Recent research shows that people usually try to avoid exerting cognitive effort yet they are willing to exert effort to gain rewards. This cost-benefit framework provides predictions for behaviour outside the laboratory. Nevertheless, the extent to which such considerations affect real-life decisions is not clear. Here we experimentally examined computer-programmers' decisions to write code in a reusable manner, using functions, which demands an initial investment of cognitive effort or to clone and tweak code, a strategy whose cost increases with repetitions. Only a small portion of our participants demonstrated sensitivity to the benefits and costs of programming strategies. When asked to solve the tasks, participants tended to avoid using functions, demonstrating biased effort estimation. By contrast, when asked how they planned to solve the tasks, participants tended to demonstrate an opposite trend, overestimating effort, in line with an injunctive norm involving the overuse of functions. In the context of real-world problems, the effect of cost-benefit considerations may therefore be limited by task-irrelevant factors. Our interdisciplinary approach may be useful in providing novel theoretical insights and in promoting cognitive-effort investments outside the laboratory.

Need for cognition does not account for individual differences in metacontrol of decision making

Humans show metacontrol of decision making, that is they adapt their reliance on decision-making strategies toward situational differences such as differences in reward magnitude. Specifically, when higher rewards are at stake, individuals increase reliance on a more accurate but cognitively effortful strategy. We investigated whether the personality trait Need for Cognition (NFC) explains individual differences in metacontrol. Based on findings of cognitive effort expenditure in executive functions, we expected more metacontrol in individuals low in NFC. In two independent studies, metacontrol was assessed by means of a decision-making task that dissociates different reinforcement-learning strategies and in which reward magnitude was manipulated across trials. In contrast to our expectations, NFC did not account for individual differences in metacontrol of decision making. In fact, a Bayesian analysis provided moderate to strong evidence against a relationship between NFC and metacontrol. Beyond this, there was no consistent evidence for relationship between NFC and overall model-based decision making. These findings show that the effect of rewards on the engagement of effortful decision-making strategies is largely independent of the intrinsic motivation for engaging in cognitively effortful tasks and suggest a differential role of NFC for the regulation of cognitive effort in decision making and executive functions.

Representational exchange in social learning: Blurring the lines between the ritual and instrumental

We propose that human social learning is subject to a trade-off between the cost of performing a computation and the flexibility of its outputs. Viewing social learning through this lens sheds light on cases that seem to violate bifocal stance theory (BST) – such as high-fidelity imitation in instrumental action – and provides a mechanism by which causal insight can be bootstrapped from imitation of cultural practices.

Task Learnability Modulates Surprise but Not Valence Processing for Reinforcement Learning in Probabilistic Choice Tasks

The goal of temporal difference (TD) reinforcement learning is to maximize outcomes and improve future decision-making. It does so by utilizing a prediction error (PE), which quantifies the difference between the expected and the obtained outcome. In gambling tasks, however, decision-making cannot be improved because of the lack of learnability. On the basis of the idea that TD utilizes two independent bits of information from the PE (valence and surprise), we asked which of these aspects is affected when a task is not learnable. We contrasted behavioral data and ERPs in a learning variant and a gambling variant of a simple two-armed bandit task, in which outcome sequences were matched across tasks. Participants were explicitly informed that feedback could be used to improve performance in the learning task but not in the gambling task, and we predicted a corresponding modulation of the aspects of the PE. We used a model-based analysis of ERP data to extract the neural footprints of the valence and surprise information in the two tasks. Our results revealed that task learnability modulates reinforcement learning via the suppression of surprise processing but leaves the processing of valence unaffected. On the basis of our model and the data, we propose that task learnability can selectively suppress TD learning as well as alter behavioral adaptation based on a flexible cost-benefit arbitration.

Valence bias in metacontrol of decision making in adolescents and young adults

The development of metacontrol of decision making and its susceptibility to framing effects were investigated in a sample of 201 adolescents and adults in Germany (12-25 years, 111 female, ethnicity not recorded). In a task that dissociates model-free and model-based decision making, outcome magnitude and outcome valence were manipulated. Both adolescents and adults showed metacontrol and metacontrol tended to increase across adolescence. Furthermore, model-based decision making was more pronounced for loss compared to gain frames but there was no evidence that this framing effect differed with age. Thus, the strategic adaptation of decision making continues to develop into young adulthood and for both adolescents and adults, losses increase the motivation to invest cognitive resources into an effortful decision-making strategy.

Plans or Outcomes: How Do We Attribute Intelligence to Others?

Humans routinely make inferences about both the contents and the workings of other minds based on observed actions. People consider what others want or know, but also how intelligent, rational, or attentive they might be. Here, we introduce a new methodology for quantitatively studying the mechanisms people use to attribute intelligence to others based on their behavior. We focus on two key judgments previously proposed in the literature: judgments based on observed outcomes (you're smart if you won the game) and judgments based on evaluating the quality of an agent's planning that led to their outcomes (you're smart if you made the right choice, even if you didn't succeed). We present a novel task, the maze search task (MST), in which participants rate the intelligence of agents searching a maze for a hidden goal. We model outcome-based attributions based on the observed utility of the agent upon achieving a goal, with higher utilities indicating higher intelligence, and model planning-based attributions by measuring the proximity of the observed actions to an ideal planner, such that agents who produce closer approximations of optimal plans are seen as more intelligent. We examine human attributions of intelligence in three experiments that use MST and find that participants used both outcome and planning as indicators of intelligence. However, observing the outcome was not necessary, and participants still made planning-based attributions of intelligence when the outcome was not observed. We also found that the weights individuals placed on plans and on outcome correlated with an individual's ability to engage in cognitive reflection. Our results suggest that people attribute intelligence based on plans given sufficient context and cognitive resources and rely on the outcome when computational resources or context are limited.

A mosaic of cost-benefit control over cortico-striatal circuitry

Dopamine contributes to cognitive control through well-established effects in both the striatum and cortex. Although earlier work suggests that dopamine affects cognitive control capacity, more recent work suggests that striatal dopamine may also impact on cognitive motivation. We consider the emerging perspective that striatal dopamine boosts control by making people more sensitive to the benefits versus the costs of cognitive effort, and we discuss how this sensitivity shapes competition between controlled and prepotent actions. We propose that dopamine signaling in distinct cortico-striatal subregions mediates different types of cost-benefit tradeoffs, and also discuss mechanisms for the local control of dopamine release, enabling selectivity among cortico-striatal circuits. In so doing, we show how this cost-benefit mosaic can reconcile seemingly conflicting findings about the impact of dopamine signaling on cognitive control.

Memory as a Computational Resource

Computer scientists have long recognized that naive implementations of algorithms often result in a paralyzing degree of redundant computation. More sophisticated implementations harness the power of memory by storing computational results and reusing them later. We review the application of these ideas to cognitive science, in four case studies (mental arithmetic, mental imagery, planning, and probabilistic inference). Despite their superficial differences, these cognitive processes share a common reliance on memory that enables efficient computation.

Hierarchical Action Control: Adaptive Collaboration Between Actions and Habits

It is now commonly accepted that instrumental actions can reflect goal-directed control; i.e., they can show sensitivity to changes in the relationship to and the value of their consequences. With overtraining, stress, neurodegeneration, psychiatric conditions, or after exposure to various drugs of abuse, goal-directed control declines and instrumental actions are performed independently of their consequences. Although this latter insensitivity has been argued to reflect the development of habitual control, the lack of a positive definition of habits has rendered this conclusion controversial. Here we consider various alternative definitions of habit, including recent suggestions they reflect chunked action sequences, to derive criteria with which to categorize responses as habitual. We consider various theories regarding the interaction between goal-directed and habitual controllers and propose a collaborative model based on their hierarchical integration. We argue that this model is consistent with the available data, can be instantiated both at an associative level and computationally and generates interesting predictions regarding the influence of this collaborative integration on behavior.

Metacontrol of decision-making strategies in human aging

Humans employ different strategies when making decisions. Previous research has reported reduced reliance on model-based strategies with aging, but it remains unclear whether this is due to cognitive or motivational factors. Moreover, it is not clear how aging affects the metacontrol of decision making, that is the dynamic adaptation of decision-making strategies to varying situational demands. In this cross-sectional study, we tested younger and older adults in a sequential decision-making task that dissociates model-free and model-based strategies. In contrast to previous research, model-based strategies led to higher payoffs. Moreover, we manipulated the costs and benefits of model-based strategies by varying reward magnitude and the stability of the task structure. Compared to younger adults, older adults showed reduced model-based decision making and less adaptation of decision-making strategies. Our findings suggest that aging affects the metacontrol of decision-making strategies and that reduced model-based strategies in older adults are due to limited cognitive abilities.

Mental labour

Mental effort is an elementary notion in our folk psychology and a familiar fixture in everyday introspective experience. However, as an object of scientific study, mental effort has remained rather elusive. Cognitive psychology has provided some tools for understanding how effort impacts performance, by linking effort with cognitive control function. What has remained less clear are the principles that govern the allocation of mental effort. Under what circumstances do people choose to invest mental effort, and when do they decline to do so? And what regulates the intensity of mental effort when it is applied? In new and promising work, these questions are being approached with the tools of behavioural economics. Though still in its infancy, this economic approach to mental effort research has already uncovered important aspects of effort-based decision-making, and points clearly to future lines of inquiry, including some intriguing opportunities presented by recent artificial intelligence research.