About

Julia Leonard is an Assistant Professor of Psychology at Yale University, holding a Ph.D. from the Massachusetts Institute of Technology obtained in 2018. Her research focuses on the fundamental cognitive, neural, and computational representations underlying children’s decisions about effort allocation. She studies children from diverse populations, ranging in age from infancy to early adolescence, employing both observational and experimental methods that integrate behavioral, computational, and neuroimaging techniques. Her ultimate goal is to gain a comprehensive and causal understanding of how children adaptively learn in response to various environmental inputs. She aims to apply her research to develop interventions that improve children’s academic, social, and health outcomes. Julia Leonard's work emphasizes understanding the effort-related decision-making processes in children, contributing to the fields of developmental psychology, neuroscience, and cognitive science.

Research topics

• Computer Science
• World Wide Web
• Applied psychology
• Psychology
• Social psychology
• Data science

Selected publications

• Children understand how adults' achievement goals drive actions

  2025-03-27

  preprintOpen accessSenior author

  Adults often hold different goals for children’s achievement: Sometimes they want a child to learn and develop their skills as much as possible (i.e., a learning goal), while other times they may forego a child’s learning in favor of successful performance (i.e., a performance goal). How do children think these achievement goals influence adults’ child-directed behaviors? Across two preregistered experiments (n = 90 adults; n = 160 5- to 8-year-old children), we found that children systematically predict that an adult would select a more difficult task for a recipient child when the adult held a learning (vs. performance) goal, and when the recipient was more competent. Importantly, we found that this pattern matched adults’ actual task choices, although adults showed more sensitivity to choosing a task that anchors closely to what a child can reasonably learn from or accomplish. These results suggest children can reason about how adult’s achievement goals manifest into observable actions, which may have consequences for children’s own goal orientations and task selections.

  PublisherOA PDFDOI

• Children understand how adults' achievement goals drive actions

  2025-03-26

  preprintOpen accessSenior author

  Adults often hold different goals for children’s achievement: Sometimes they want a child to learn and develop their skills as much as possible (i.e., a learning goal), while other times they may forego a child’s learning in favor of successful performance (i.e., a performance goal). How do children think these achievement goals influence adults’ child-directed behaviors? Across two preregistered experiments (n = 90 adults; n = 160 5- to 8-year-old children), we found that children systematically predict that an adult would select a more difficult task for a recipient child when the adult held a learning (vs. performance) goal, and when the recipient was more competent. Importantly, we found that this pattern matched adults’ actual task choices, although adults showed more sensitivity to choosing a task that anchors closely to what a child can reasonably learn from or accomplish. These results suggest children can reason about how adult’s achievement goals manifest into observable actions, which may have consequences for children’s own goal orientations and task selections.

  PublisherOA PDFDOI

• When Bayesians take over: A computational model of parental intervention

  2025-05-26

  preprintOpen accessSenior author

  When children encounter challenges, parents often wonder: Should I let my child figure it out or take over? How parents resolve this dilemma shapes key developmental outcomes, yet we know little about the cognitive mechanisms that drive these decisions. Here, we model parental "take over" decisions as a Bayesian solution to a Partially Observable Markov Decision Process (POMDP) and qualitatively compare model predictions with behavioral data from parent-child interactions. We find that two core beliefs guide intervention: the child’s probability of success and the utility of the task. Parents are more likely to take over when they believe their child is less skilled and the task is harder, and more likely to step back when they expect the rewards of independent effort to outweigh the costs. The model captures how these beliefs interact to shape decision-making and, together with the empirical data, reveals the cognitive computations that underlie parental intervention.

  PublisherOA PDFDOI

• Exploration is associated with socioeconomic disparities in learning and academic achievement in adolescence

  Nature Communications · 2025-07-09 · 4 citations

  articleOpen access

  Adolescents from lower socioeconomic status backgrounds often underperform on tests of learning and academic achievement. Existing theories propose that these disparities reflect not only external constraints, like limited resources, but also internal decision strategies that adapt to the early environment and influence learning. These theories predict that adolescents from lower socioeconomic status backgrounds explore less and exploit more, which, in turn, reduces learning and academic achievement. Here, we test this possibility and show that lower socioeconomic status in adolescence is associated with less exploration on a reward learning task (n = 124, 12-14-year-olds from the United States). Computational modeling revealed that reduced exploration was related to higher loss aversion. Reduced exploration also mediated socioeconomic differences in task performance, school grades, and, in a lower-socioeconomic status subsample, academic skills. These findings raise the possibility that learning disparities across socioeconomic status relate not only to external constraints but also to internal decision strategies and provide some mechanistic insight into the academic achievement gap.

  PublisherOA PDFDOI

• Children Predict Improvement on Novel Skill Learning Tasks

  Child Development · 2025-04-04 · 3 citations

  articleSenior author

  Learning takes time: Performance usually starts poorly and improves with practice. Do children intuit this basic phenomenon of skill learning? In preregistered Experiment 1 (n = 125; 54% female; 48% White; collected 2022-2023), US 7- to 8-year-old children predicted improved performance, 5- to 6-year-old children predicted flat performance, and 4-year-old children predicted near-instant success followed by worse performance on a novel skill learning task. In preregistered Experiment 2 (n = 75; 47% female; 69% White; collected 2023), on a task with lowered cognitive demands, US 4- to 6-year-old children predicted improved performance. Thus, although children expect to improve on novel tasks, younger children need scaffolding to form these predictions and grasp this fundamental aspect of learning.

  PublisherDOI

• People accurately predict the shape but not the parameters of skill learning curves

  Cognition · 2025-02-20 · 2 citations

  articleSenior author
  PublisherDOI

• Detecting social biases using mental state inference.

  Journal of Personality and Social Psychology · 2025-06-05 · 1 citations

  articleOpen access

  = 876 total), show that this model captures participants' inferences about an agent's prior beliefs (Experiment 1), general social biases (Experiment 2) across various real-world contexts (Experiments 3a-3c), and even specific racial and gender biases (Experiment 4). We compare this model with alternative models that differ in their assumptions about whether and how a biased agent updates their beliefs about an individual. Participants' judgments were best explained as the process of inferring an agent's prior beliefs before updating them based on available evidence about the individual. These findings highlight the role of mental state reasoning in bias detection and broaden our understanding of the human capacity to detect and reason about social biases. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

  PublisherDOI

• U.S. Children Expect and Approve of Adults’ Gender Stereotypes

  2025-05-08

  preprintOpen accessSenior author

  Gender stereotypes, such as the idea that boys are more interested than girls in STEM,contribute to gender disparities in STEM. Do young children generally expect adults to holdsuch stereotypes, or do they withhold these assumptions without sufficient proof? Acrossfour preregistered experiments (n = 574), we found evidence for the former: 5- to 7-year-oldchildren in the United States predicted that teachers meeting their students for the first timewould assign engineering activities to boys and reading activities to girls, despite childrenknowing that the students liked both equally. Further, children approved of teachers assigningengineering activities only to boys and reading activities only to girls, more so than the reverse.Despite ongoing efforts to challenge gender disparities in education, our results reveal thatchildren enter school expecting teachers to hold gender stereotypes and view this as acceptable,highlighting a new early obstacle to educational equity.

  PublisherOA PDFDOI

• People accurately predict the shape but not the parameters of skill learning curves

  2025-02-05

  preprintOpen accessSenior author

  Decades of research have shown that skill learning often unfolds exponentially — people improve rapidly early on, and then performance gradually levels off. Given how important expectations of learning are for actual learning, we explored whether people accurately intuit this canonical time course of skill learning. Across six preregistered experiments (n = 500), we find that people correctly predict that skill learning curves (error reductions over time) on a novel visuomotor task will follow an exponential decay function, both for an imagined naïve player and for themselves, before engaging with the task. Moreover, people are sensitive to conditions that merit exponential learning within a bounded time frame and only predict these curves when an imagined player puts in effort and the task is not too difficult. However, people systematically misestimate specific parameters of skill learning (e.g., initial and average performance, and rate of improvement), which relates to reduced affect at the beginning of learning. Critically, these negative effects can be ameliorated by practice: Providing people with minimal practice reduces their prediction errors and, in turn, buffers them from negative feelings at the beginning of learning.

  PublisherDOI

• When Bayesians take over: A computational model of parental intervention

  2025-02-25 · 1 citations

  preprintOpen accessSenior author

  When children encounter new challenges, parents often ask themselves: should I let my child figure it out, or should I step in and do it for them? How parents resolve this dilemma is linked to various child developmental outcomes, yet we know little about the cognitive computations that underlie parents’ decision to intervene. Here, we model parenting decisions as a Bayesian solution to a Partially Observable Markov Decision Process (POMDP) and qualitatively compare predictions with behavioral data from real-time parent-child interactions. Empirically, we find that parents are more likely to take over when they believe children are less skilled and when tasks are challenging, and more likely to step back when they believe their child can learn from doing the task on their own. The model captures the fine-grained ways in which these factors shape parent decision-making and, alongside the empirical data, uncovers the cognitive computations that drive parental intervention.

  PublisherOA PDFDOI

Frequent coauthors

• John D. E. Gabrieli

  Massachusetts Institute of Technology

  120 shared

• Allyson P. Mackey

  California University of Pennsylvania

  91 shared

• Amy S. Finn

  49 shared

• Rachel Romeo

  University of Maryland, College Park

  45 shared

• Sydney T. Robinson

  Philadelphia University

  33 shared

• Rebecca Martin

  University of Colorado Denver

  24 shared

• Martin R. West

  22 shared

• Alison Gopnik

  University of California, Berkeley

  18 shared