Most recent research on human planning attempts to adjudicate between a small set of hypothesized models based on their ability to predict participants' choices, using carefully designed experiments and/or model comparison. Here, we propose an alternative approach. We designed a task in which gaze is highly indicative of participants' planning operations, allowing us to discover properties of human planning from eye-tracking data in a data-driven way. Our results reveal ways that people's planning strategies have both similarities and differences with classical planning algorithms like best-first search and Monte Carlo tree search. They also provide a more nuanced perspective on previously proposed properties of human planning like pruning and depth limits. We conclude that planning research would benefit greatly from an increased use of rich sources of data that provide more direct evidence about the internal processes underlying sequential decision-making.

Across the lifespan, humans direct their learning towards information they are curious to know. However, it is unclear what elicits curiosity, and whether and how this changes across development. Is curiosity triggered by surprise and uncertainty, as prior research suggests, or by expected learning, which is often confounded with these features? In the present research, we use a Bayesian reinforcement learning model to quantify and disentangle surprise, uncertainty, and expected learning. We use the resulting model-estimated features to predict curiosity ratings from 5- to 9-year-olds and adults in an augmented multi-armed bandit task. Like adults’ curiosity, children’s curiosity was best predicted by expected learning. However, after accounting for expected learning, children (but not adults) were also more curious when uncertainty was higher and surprise lower. This research points to developmental changes in what elicits curiosity and calls for a reexamination of research that confounds these elicitors.

How can people learn to make better decisions and be-come more far-sighted? To make the underlying learningmechanisms more accessible to scientific inquiry, we developa computational method for measuring the time course ofexperience-dependent changes in people’s planning strategies.We validated our method on simulated and empirical data: onsimulated data its inferences were significantly more accuratethan simpler approaches, and when evaluated on human datait correctly detected the plasticity-enhancing effect of perfor-mance feedback. Having validated our method, we illustratehow it can be used to gain new insights into the time courseand nature of cognitive plasticity. Future work will leverageour method to i) reverse-engineer the learning mechanismsenabling people to acquire complex cognitive skills such asplanning and problem-solving and ii) measure individual dif-ferences in cognitive plasticity.

Curiosity motivates exploration and is beneficial for learning,but curiosity is not always experienced when facing theunknown. In the present research, we address this selectivity:what causes curiosity to be experienced under somecircumstances but not others? Using a Bayesian reinforcementlearning model, we disentangle four possible influences oncuriosity that have typically been confounded in previousresearch: surprise, local uncertainty/expected informationgain, global uncertainty, and global expected informationgain. In two experiments, we find that backward-lookinginfluences (concerning beliefs based on prior experience) andforward-looking influences (concerning expectations aboutfuture learning) independently predict reported curiosity, andthat forward-looking influences explain the most variance.These findings begin to disentangle the complexenvironmental features that drive curiosity.

Working memory is widely assumed to underlie multi-step planning, where representations of possible future actions and rewards are iteratively updated before determining a choice. But most working memory research focuses on a context where stimuli are presented simultaneously and the value of encoding each stimulus is independent of others. It is unclear how working memory functions in planning scenarios where the rewards of future actions unfold over time, are retained in working memory, and must be integrated for plan selection. To bridge this gap, we adapted a version of the "mouselab task" in which participants sequentially observe the reward at each node in a decision tree before selecting a plan that maximizes cumulative rewards. We specified a theoretical model to characterize the optimal encoding and maintenance strategy for this task given the working memory constraints, which trades off the cost of storing information with the potential benefit of informing later choices. The model encoded rewards in choice-relevant plans more often, in particular, rewards on the best and (to a lesser extent) worst plans. We then tested this hypotheses on human participants, who showed the same pattern in the accuracy of their explicit recall. Our study thus establishes an empirical and theoretical foundation for models of how people encode and maintain information during planning.

When faced with a complex problem, people naturally break it up into several simpler problems. This hierarchical decom-position of an ultimate goal into sub-goals facilitates planning by reducing the number of factors that must be consideredat one time. However, it can also lead to suboptimal decision-making, obscuring opportunities to make progress towardsmultiple subgoals with a single action. Is it possible to take advantage of the hierarchical structure of problems withoutsacrificing opportunities to kill two birds with one stone? We propose that people are able to do this by representing andpursuing multiple subgoals at once. We present a formal model of planning with compositional goals, and show that itexplains human behavior better than the standard ”one-at-a-time” subgoal model as well as non-hierarchical limited-depthsearch models. Our results suggest that people are capable of representing and pursuing multiple subgoals at once; how-ever, there are limitations on how many subgoals one can pursue concurrently. We find that these limitations vary byindividual.

People’s judgments and decisions often deviate from classicalnotions of rationality, incurring costs to both themselves and tosociety. Previous research has proposed that the cost of thesebiases can be reduced by redesigning decision problems basedon theories of human decision making. These modifications—or nudges—can have dramatic results and have been success-fully applied to variety of domains. However, the formal un-derpinning of nudge theory is limited, and it is not always clearwhat the effect of a nudge will be before it is implemented. Asa result, designing nudges can be time consuming and error-prone. In this paper, we propose an automatic method for de-riving optimal nudges. The method is based on a resource-rational model, which assumes that people make decisions ina way that achieves a near-optimal tradeoff between the costand benefits of deliberation. We then frame nudges as modi-fications to the costs of different cognitive operations, encour-aging the cognitively frugal decision maker to consider someproblem features over others. As a proof of concept, we applythe method to the Mouselab process-tracing paradigm, findingthat optimal nudges lead participants to make better decisionswith less cognitive effort.

In the history of cognitive science, there have been two com-peting philosophies regarding how people reason about theworld. In one, people rely on rich, generative models to makepredictions about a wide range of scenarios; while in the other,people have a large “bag of tricks”, idiosyncratic heuristics thattend to work well in practice. In this paper, we suggest thatrather than being in opposition to one another, these two ideascomplement each other. We argue that people’s capacity formental simulation may support their ability to learn new cue-based heuristics, and demonstrate this phenomenon in two ex-periments. However, our results also indicate that participantsare far less likely to learn a heuristic when there is no logical orexplicitly conveyed relationship between the cue and the rele-vant outcome. Furthermore, simulation—while a potentiallyuseful tool—is no substitute for real world experience

Research in psychology and artificial intelligence has sought to ground information-seeking behavior in rational terms, typically assuming that people or agents prefer more informative data over less informative data. While this seems reasonable on its surface, it assumes that informativeness is only a property of the data, rather than a joint property of the data and a (potentially bounded) learner. That is, to the extent that it is hard to draw the right inferences from data that are theoretically "high information," the data will not actually be highly informative to the learner. Here, we investigate active learning in humans using the code-breaking game Mastermind, which requires deductive reasoning from evidence. We find that people make queries that are less informative than random guesses, challenging standard rational or resource-rational accounts of information-seeking. We then show that people make queries are informative to them assuming they have a bounded capacity to draw inferences. We also find that participants prefer queries that provide easily-interpretable information over queries that provide more information but are less interpretable. Our results suggest that people are aware of their own cognitive limitations and seek information that they can use.

People’s cognitive strategies are jointly shaped by function and computational constraints. Resource-rational analysis lever- ages these constraints to derive rational models of people’s cognitive strategies from the assumption that people make rational use of limited cognitive resources. We present a resource-rational analysis of planning and evaluate its predic- tions in a newly developed process tracing paradigm. In Ex- periment 1, we find that a resource-rational planning strategy predicts the process by which people plan more accurately than previous models of planning. Furthermore, in Experiment 2, we find that it also captures how people’s planning strategies adapt to the structure of the environment. In addition, our ap- proach allows us to quantify for the first time how close peo- ple’s planning strategies are to being resource-rational and to characterize in which ways they conform to and deviate from optimal planning.