About

Angela Radulescu is an Assistant Professor at the Center for Computational Psychiatry at the Icahn School of Medicine at Mount Sinai. Her research focuses on the intersection of data science and psychiatry, contributing to the development of computational models to better understand mental health disorders. She is part of a community of faculty fellows at the NYU Center for Data Science, which supports research across academia, industry, and government, emphasizing interdisciplinary collaboration and innovative applications of data science.

Research topics

• Artificial Intelligence
• Computer Science
• Cognitive science
• Psychology
• Cognitive psychology
• Knowledge management
• Neuroscience
• Zoology
• Epistemology
• Internal medicine
• Social psychology
• Immunology
• Pediatrics
• Medicine
• Endocrinology

Selected publications

• The roots of rationality

  2026-04-12

  articleOpen access

  Functional accounts of mind, brain, and behavior have profoundly shaped thinking across many disciplines. Despite this common lens, the processes driving adaptive outcomes are often studied in isolation, leading to narrow and localized explanations. This paper brings together perspectives across fields by identifying six dynamics that enable agents to behave adaptively. These dynamics span multiple timescales and systems, including individual-level processes—thinking, learning, and development—and population-level processes—market selection, cultural evolution, and genetic evolution. We examine the reasons behind functional explanations of cognition and behavior, the conditions that foster rationality, and the interactions between adaptive processes. Our goal is to bring the cognitive, social, biological, and computer sciences into closer conversation by exposing researchers to a wide range of perspectives on adaptive decision-making.

  PublisherDOI

• A Resource-Rational Account of Human Eye Movements During Immersive Visual Search

  Open Mind · 2026-01-01

  articleOpen access1st authorCorresponding

  The nature of eye movements during visual search has been widely studied in cognitive science. Virtual reality (VR) paradigms are an opportunity to test whether computational models of search can predict naturalistic search behavior. However, existing ideal observer models are constrained by strong assumptions about the structure of the world, rendering them impractical for modeling the complexity of environments which can be studied in VR. To address these limitations, we modeled immersive visual search as a reinforcement learning problem, in which sequential decisions are made over a multidimensional representation of the environment learned by a convolutional neural network. In our formulation, RL agents learned a policy over latent states-effectively solving what is known as a meta-Markov decision process (meta-MDP), where each decision concerns how to allocate attention to information in the environment. Training deep-RL agents on the meta-MDP showed that learned (i.e., optimal) search policies converge to a classic ideal-observer model of search developed for simple (1D) stimuli. We compared the learned resource-rational policy with human gaze data from a visual-search experiment conducted in VR and found qualitative and quantitative alignment between model predictions and human behavior. However, both the model's simulated performance and its correspondence with human behavior depended strongly on the representational features available to the policy. These results suggest that naturalistic visual search behavior can partially be explained by resource-rational allocation of limited cognitive resources, and the choice of representation influences the degree of alignment between model and human behavior.

  PublisherDOI

• Resolution of Hashimoto thyroiditis with Janus kinase inhibitor therapy in a patient with alopecia universalis

  JCEM Case Reports · 2026-03-02

  articleOpen accessSenior author

  Hashimoto thyroiditis (HT) is the most common autoimmune thyroid disorder, marked by lymphocytic inflammation and progressive hypothyroidism. Its pathophysiology involves both T-cell-mediated tissue destruction and the production of autoantibodies against thyroid peroxidase (TPO) and thyroglobulin, with TPO antibodies present in over 90% of cases. Standard treatment with thyroid hormone replacement can leave residual symptoms, highlighting the need for disease-modifying therapies. Recent advances in immunotherapy have identified Janus kinase inhibitors, such as tofacitinib and baricitinib, as promising agents for autoimmune conditions. These drugs target the Janus kinase/signal transducer and activator of transcription pathway, which mediates pro-inflammatory cytokines implicated in HT, including interferon-γ and interleukin-6. We report a case of a 44-year-old woman with both alopecia universalis and HT who experienced normalization of TPO antibody levels, no longer requiring thyroid hormone treatment following JAK inhibitor therapy for alopecia universalis-suggesting possible reversal of HT. This case highlights the immunomodulatory potential of JAK inhibitors in HT and supports further investigation into their therapeutic role in addressing both hormonal and autoimmune mechanisms in thyroid disease.

  PublisherDOI

• The roots of rationality

  PsyArXiv (OSF Preprints) · 2026-04-11

  preprintOpen access

  Functional accounts of mind, brain, and behavior have profoundly shaped thinking across many disciplines. Despite this common lens, the processes driving adaptive outcomes are often studied in isolation, leading to narrow and localized explanations. This paper brings together perspectives across fields by identifying six dynamics that enable agents to behave adaptively. These dynamics span multiple timescales and systems, including individual-level processes—thinking, learning, and development—and population-level processes—market selection, cultural evolution, and genetic evolution. We examine the reasons behind functional explanations of cognition and behavior, the conditions that foster rationality, and the interactions between adaptive processes. Our goal is to bring the cognitive, social, biological, and computer sciences into closer conversation by exposing researchers to a wide range of perspectives on adaptive decision-making.

  Publisher

• Learning to Attend Through Value-Based Hypothesis Testing

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-03-12

  articleOpen accessSenior authorCorresponding

  In complex environments, humans must determine which features are relevant for learning and decision-making. Psychological theories offer competing accounts of this process: associative models suggest that attention emerges gradually through learned changes in feature values, whereas hypothesis-driven accounts propose that learners selectively attend to actively tested rules. Because attentional states are covert, similar behavior can arise from different underlying strategies, making these accounts difficult to distinguish using choice data alone. We inferred latent attention dynamics during learning and decision-making by training recurrent neural networks on synthetic data generated from feature-based reinforcement learning (FRL) and serial hypothesis testing (SHT) models. A network trained on hybrid (FRL+SHT) data outperformed single-model networks, decoding latent human attention with more than 80% accuracy. These results suggest that human attention reflects an interaction between value-based learning and hypothesis testing, in which learned feature value guides the generation and evaluation of candidate rules.

  PublisherOA PDFDOI

• Rewards transiently and automatically enhance sustained attention.

  Journal of Experimental Psychology General · 2025-01-21 · 2 citations

  articleOpen access

  = 135), we observed that even passively received rewards elicit transient boosts in sustained attention. Together, these findings suggest that rewards transiently buffer against attentional lapses to improve vigilance, likely through generic increases in arousal or motivation. These results point to a fundamental relationship between reward and sustained attention. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

  PublisherDOI

• 256. Investigating the Neural Encoding of Reward Prediction Errors and Anhedonia Using 7-Tesla MRI

  Biological Psychiatry · 2025-04-09

  article
  PublisherDOI

• VTA network dominance in depression confers distinct psychopathological states through blunted neural tracking of reward prediction errors

  Research Square · 2025-09-09

  preprintOpen accessSenior author
  PublisherOA PDFDOI

• Unveiling value functions in social cognition with multi-agent inverse reinforcement learning

  bioRxiv (Cold Spring Harbor Laboratory) · 2024-10-12 · 2 citations

  preprintOpen access

  Abstract Social behavior requires individuals to consider not only their own goals but also those of others. Latent value functions that encode such goals can be recovered from behavior using inverse reinforcement learning in single-agent settings. However, extending it to multi-agent interactions is challenging, because value functions are defined over joint state spaces that grow exponentially with the number of agents. Existing approaches often manage this complexity by imposing strong structural assumptions about social interactions, thereby limiting their applicability and interpretability. Here we show that joint value functions governing social interactions can be effectively represented through value decomposition into individual value maps for each agent and low-dimensional interaction terms. We develop a multi-agent inverse reinforcement learning framework (MAIRL) to infer these representations from behavior. In mouse and primate social tasks, MAIRL reveals interpretable value maps that are conditioned on the distinct social roles animals play during group behavior. Together, these results establish MAIRL as an interpretable and scalable framework for identifying latent value representations guiding multi-agent behavior across species.

  PublisherOA PDFDOI

• Emotional Overshadowing: Pleasant and Unpleasant Cues Overshadow Neutral Cues in Human Associative Learning

  Affective Science · 2024-09-01 · 1 citations

  articleOpen access

  Abstract When learning about stimuli comprised of multiple cues, humans and other animals tend to form stronger cue-outcome associations for more salient cues than for less salient cues. This phenomenon, termed overshadowing , has typically been demonstrated between cues that vary in salience because of differences in physical intensity. In this study, we investigated whether differences in the emotional valence of cues in a compound stimulus similarly led to differences in the strength of cue-outcome learning. Using a probabilistic categorisation task in which stimuli were compounds consisting of pairs of emotional or non-emotional cue images , we found consistent evidence for emotional overshadowing across both an initial exploratory study ( N = 50) and a confirmatory preregistered replication study ( N = 200). Specifically, both pleasant and unpleasant cue images tended to overshadow neutral cue images, but pleasant and unpleasant cue images did not overshadow one another. Moreover, across stimuli, the magnitude of differences in learning between cues was proportional to differences in their absolute emotional valence, suggesting that attentional capture by both positively and negatively valenced emotions drives overshadowing. These findings have implications for understanding associative learning in natural environments, where stimuli are frequently imbued with emotional valence prior to learning.

  PublisherOA PDFDOI

Frequent coauthors

• Yael Niv

  Princeton University

  24 shared

• Daniel Bennett

  Monash University

  7 shared

• Lynn A. Burmeister

  University of Minnesota

  6 shared

• Frank Q. Nuttall

  University of Minnesota

  6 shared

• Mary C. Gannon

  Minneapolis VA Health Care System

  6 shared

• T. Araki

  University of Minnesota Medical Center

  5 shared

• Bas van Opheusden

  5 shared

• Reka Daniel

  4 shared

Labs

• Icahn School of Medicine at Mount SinaiPI

Awards & honors

• CDS Faculty Fellows and Moore-Sloan Fellows at CDS
• Alumni of these programs now hold faculty positions at Yale,…
• Vincent Divol: AI Fellow, Universite PSL
• Najoung Kim: Assistant Professor, Boston University
• Byung-Doh Oh: Assistant Professor, Nanyang Technological Uni…