About

Michael J. Frank, PhD, is a Professor in the Department of Cognitive and Psychological Sciences at Brown University and serves as the Neuroscience Graduate Program Director. He is affiliated with the Nancy G Zimmerman Center for Computational Brain Science and the Carney Institute for Brain Science, as well as the Department of Psychiatry and Human Behavior. His research integrates computational modeling and experimental approaches to investigate the neural mechanisms underlying reinforcement learning, decision making, and cognitive control. Specifically, his work focuses on developing neural circuit and algorithmic models that simulate interactions between brain areas such as the prefrontal cortex and basal ganglia, with an emphasis on dopamine modulation. These models are tested through neuropsychological, pharmacological, genetic, and imaging techniques, primarily EEG. Michael Frank's educational background includes a PhD in Neuroscience and Psychology from the University of Colorado at Boulder, where he was part of the Computational Cognitive Neuroscience Lab, an MS in Electrical and Computer Engineering with a biomedicine option from the same institution, and a BSc in Electrical Engineering from Queen's University in Canada. He teaches courses related to computational cognitive neuroscience and motivated decision making, reflecting his expertise in computational approaches to understanding brain function and behavior.

Research topics

• Artificial Intelligence
• Computer Science
• Psychology
• Neuroscience
• Cognitive psychology
• Clinical psychology
• Social psychology
• Mathematics
• Systems engineering
• Engineering
• Data science

Selected publications

• Pitavastatin effects on lipids in relation to major adverse cardiovascular events: a REPRIEVE secondary analysis

  The Lancet HIV · 2026-04-03 · 1 citations

  articleOpen access

  BACKGROUND: The Randomized Trial to Prevent Vascular Events in HIV (REPRIEVE) found a 36% reduction in major adverse cardiovascular events (MACE) with pitavastatin in people with HIV. Little is known about the relationship between lipid lowering and MACE in this population. We evaluated pitavastatin effects on lipids and examined mediation of pitavastatin effects on MACE through lipid lowering. METHODS: REPRIEVE was a randomised, double-blind, placebo-controlled phase 3 trial evaluating pitavastatin (4 mg daily) or placebo for prevention of MACE in people with HIV. Participants aged 40-75 years, on stable combination antiretroviral therapy (ART), and at low-to-moderate atherosclerotic cardiovascular disease risk with minimally elevated LDL were followed up for a median of 5·6 years. Centrally tested fasting lipids were captured at entry and annually thereafter. The primary MACE outcome, reported previously, was time to first MACE. Here, we report secondary outcomes on fasting lipids. Linear mixed-effects models were used for assessment of the pitavastatin effect on lipids, Cox regression for relationship of lipids to MACE, and Vansteelandt method for mediation analysis. FINDINGS: REPRIEVE enrolled 7769 participants from March 26, 2015, to July 31, 2019, in 12 countries across five Global Burden of Diseases super-regions. The median baseline LDL was 108 mg/dL, and similar across treatment groups. Pitavastatin effects were primarily observed on LDL, with a modest reduction in triglycerides and no apparent effect on HDL. Based on the longitudinal data modelling, the estimated treatment group difference in LDL at month 12 (pitavastatin minus placebo) was -30 mg/dL (95% CI -31 to -29), corresponding to a 30% reduction. The estimated risk of LDL of ≥100 mg/dL at month 12 was 0·18 in the pitavastatin group and 0·57 in the placebo group (relative risk 0·32, 95% CI 0·30-0·34). A 30% lower time-updated average LDL was associated with 20% lower risk of primary MACE (hazard ratio 0·80, 95% CI 0·68-0·94). Of the pitavastatin effect on MACE, 68% was estimated to be mediated through LDL, although with low precision (95% CI 15-574). INTERPRETATION: LDL is strongly related to MACE, and LDL lowering should be an important goal of primary cardiovascular prevention in people with HIV, even in those with minimally elevated LDL. Treatment should aim to achieve accepted primary care prevention targets for LDL. FUNDING: US National Institutes of Health, Kowa Pharmaceuticals America, Gilead Sciences, and ViiV Healthcare.

  PublisherOA PDFDOI

• A Novel Approach-Avoidance Task to Study Decision Making Under Outcome Uncertainty

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-07-17

  preprintOpen access

  To behave adaptively, people need to integrate information about probabilistic outcomes and balance drives to approach positive outcomes and avoid negative outcomes. However, questions remain about how uncertainty in positive and negative outcomes influence approach-avoid decision-making dynamics. To fill this gap, we developed a novel Probabilistic Approach Avoidance Task (PAAT) and characterized behavior in this task using sequential sampling models. In this task, participants (N=34, 24 females) made a series of choices between pairs of options, each consisting of variable probabilities of reaching a positive outcome (monetary reward) and of reaching a negative outcome (aversive image). Participants tended to choose options that maximized the likelihood of reward and minimized the likelihood of aversive outcomes. Moreover, the weights they placed on each of these differed for choices where these likelihoods were in opposition (i.e., the riskier option was also more rewarding; incongruent trials) relative to when these were aligned (congruent trials). Computational modeling revealed that the relative influence of rewarding and aversive outcomes on choice was captured by differences in the rate of decision-relevant information accumulation. These modeling results were validated with a series of model comparisons and posterior predictive checks, demonstrating that our sequential sampling models reliably captured our behavioral data. Together, these findings improve our understanding of the influence of motivational conflict, outcome type, and levels of uncertainty on approach-avoid decision-making.

  PublisherOA PDFDOI

• eLife Assessment: Reduced discrimination between signals of danger and safety but not overgeneralization is linked to exposure to childhood adversity in healthy adults

  2025-02-20

  peer-reviewOpen access1st authorCorresponding

  Childhood adversity is associated with aberrant threat learning patterns in a large healthy adult sample as evidenced by empirically testing several theories linking childhood adversity to psychopathology.

  PublisherDOI

• Model-Based Electroencephalography Phenotyping Uncovers Distinct Neurocomputational Mechanisms Underlying Learning Impairments Across Psychopathologies

  Biological Psychiatry Global Open Science · 2025-11-29

  articleOpen accessSenior author

  Major depressive disorder (MDD), bipolar disorder (BP), and schizophrenia (SCZ) involve learning impairments with poorly understood mechanisms. Understanding both the similarities and differences in these mechanisms is important to guide the development of new, targeted interventions. 255 participants diagnosed with MDD (n=54), BP (n=47), SCZ (n=67) or without any diagnoses (CTRL; n=87) performed an associative learning task. Computational modeling quantified the mechanistic interplay between working memory (WM) and reinforcement learning (RL). The latent RL and WM signatures in the EEG dynamics showed shared and distinct neurocognitive mechanisms underlying learning. All clinical groups showed learning impairments at the behavioral level. Model-based EEG analyses linked these impairments to distinct patterns in the dynamic interplay between latent RL and WM mechanisms, contrasting with the typical patterns observed in CTRL. SCZ was characterized by reduced neural markers of WM, weakening the cooperative influence of WM onto RL (“reduced WM recruitment”), and reduced integration of negative feedback. Conversely, MDD was characterized by reduced reciprocal influence of RL onto WM, reducing the tendency to upregulate WM contribution with reward history (“impaired WM management”). Finally, BP was characterized by deficits in both WM and RL recruitment, along with higher WM decay. Behavioral learning impairments that appear similar across clinical groups can be linked to distinct neurocognitive mechanisms via integrative neurocomputational modeling. Our approach provides insights into the interplay of underlying learning mechanisms and how they manifest differently across psychopathologies. People with depression, bipolar disorder, and schizophrenia often show learning difficulties but the underlying causes may differ. By combining brain activity recordings with computational models, we identified distinct cognitive mechanisms driving these impairments. Our findings show how modeling and physiology can give insights into hidden decision dynamics behind learning difficulties. We also outline key steps needed to advance computational psychiatry tools toward clinical applications, including their potential use in guiding personalized treatment.

  PublisherDOI

• Parallel trade-offs in human cognition and neural networks: The dynamic interplay between in-context and in-weight learning

  Proceedings of the National Academy of Sciences · 2025-08-28 · 6 citations

  articleOpen accessSenior authorCorresponding

  Human learning embodies a striking duality: Sometimes, we can rapidly infer and compose logical rules, benefiting from structured curricula (e.g., in formal education), while other times, we rely on an incremental approach or trial-and-error, learning better from curricula that are randomly interleaved. Influential psychological theories explain this seemingly conflicting behavioral evidence by positing two qualitatively different learning systems-one for rapid, rule-based inferences (e.g., in working memory) and another for slow, incremental adaptation (e.g., in long-term and procedural memory). It remains unclear how to reconcile such theories with neural networks, which learn via incremental weight updates and are thus a natural model for the latter, but are not obviously compatible with the former. However, recent evidence suggests that metalearning neural networks and large language models are capable of in-context learning (ICL)-the ability to flexibly infer the structure of a new task from a few examples. In contrast to standard in-weight learning (IWL), which is analogous to synaptic change, ICL is more naturally linked to activation-based dynamics thought to underlie working memory in humans. Here, we show that the interplay between ICL and IWL naturally ties together a broad range of learning phenomena observed in humans, including curriculum effects on category-learning tasks, compositionality, and a trade-off between flexibility and retention in brain and behavior. Our work shows how emergent ICL can equip neural networks with fundamentally different learning properties that can coexist with their native IWL, thus offering an integrative perspective on dual-process theories of human cognition.

  PublisherOA PDFDOI

• Behavioral Opportunism and Altered Dopamine Dynamics in Mice Exposed to Early Life Adversity

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-12-08

  articleOpen access

  Early life adversity (ELA) confers risk for reward-related psychopathologies. These risks may stem from adaptations optimizing reward pursuit in anticipation of unreliable, resource poor environments. One rational adaptation to poor, unreliable environments is Behavioral Opportunism: updating expectations more slowly and acting vigorously only when reward is immediately available. To systematically test the impact of ELA on behavioral strategies and underlying reward processing mechanisms, we exposed mice to resource restriction (limited bedding and nesting materials for 7 days) to manipulate the reliability and quality of early life care. Subsequently, we tested adults' reward learning and decision making in a two-arm bandit task and recorded dopamine signaling using dLight1.2 fiber photometry in the nucleus accumbens core. Exposure to ELA led to poorer choice discrimination, impaired learning, and decreased adaptation to changes in reward availability. Furthermore, ELA mice were slower to choose between levers but were faster to retrieve immediately available rewards when delivered, consistent with a strategy of behavioral opportunism. Dopamine signaling predicted behavior in both rearing conditions, and its fluctuations were strongly predictive of faster retrieval in ELA mice and an increased likelihood of choice repetition, implying that aberrant dopamine signals underlie slowed learning and vigorous action for immediately available rewards. To understand key features of maternal interactions driving these effects, we used home cage video monitoring to quantify maternal behaviors, continuously, across early life. We found that specific experiential outcomes, such as maternal kicking, intensified behavioral opportunism in adults, predicting poorer bandit task performance beyond the group effect of ELA. Behavioral opportunism provides an explanatory framework for interpreting altered reward processing and reward pursuit in adulthood for individuals exposed to ELA.

  PublisherOA PDFDOI

• Increased Reporting of Speech in Degraded Stimuli in Schizophrenia: A Case Control Study with Sine-Wave-Speech

  Schizophrenia Bulletin · 2025-12-13

  articleOpen access

  BACKGROUND AND HYPOTHESIS: People who hear voices may have strong prior expectations of speech, so that noisy auditory signals are resolved as speech. Data in non-clinical voice hearers suggest that voice hearing may involve sensitivity to speech in degraded stimuli. This has yet to be examined in people with schizophrenia (SZ). STUDY DESIGN: In this case-control study, we presented sine-wave-speech (SWS; made by replacing the formants in speech with pure tone whistles) to people with SZ (n = 63) and healthy controls (HC; n = 27). SWS is typically unintelligible on first exposure. However, once the listener knows that it is potentially intelligible as speech (by exposure to the unaltered speech template, which thus serves as a prior expectation), relatively high levels of comprehension are achieved. Our participants first listened to intelligible and unintelligible SWS and reported whether they heard speech. They were then exposed to the speech templates, and then the first phase was repeated. STUDY RESULTS: Compared to HC, people with SZ reported hearing more speech before template exposure. The Reveal increased both groups' false alarms and reporting of speech, but there was no interaction with group. Change in hit rates after the Reveal correlated with hallucinations, which is consistent with a greater influence of the priors enhancement in SZ patients who hear voices. CONCLUSIONS: These findings suggest that people with SZ have stronger expectations of speech. This task has validity for hallucinatory voice hearing. It is also simple and convenient to administer, and may prove useful in detecting prodromal risk, as well as acute exacerbation in voice hearing.

  PublisherOA PDFDOI

• Striatal dopamine can enhance both fast working memory, and slow reinforcement learning, while reducing implicit effort cost sensitivity

  Nature Communications · 2025-07-09 · 5 citations

  articleOpen access

  Abstract Associations can be learned incrementally, via reinforcement learning (RL), or stored instantly in working memory (WM). While WM is fast, it is also capacity-limited and effortful. Striatal dopamine may promote WM, by facilitating WM updating and effort exertion and also RL, by boosting plasticity. Yet, prior studies have failed to distinguish between the effects of dopamine manipulations on RL versus WM. N = 100 participants completed a paradigm isolating these systems in a double-blind study measuring dopamine synthesis with [ 18 F]-FDOPA PET imaging and manipulating dopamine with methylphenidate and sulpiride. We find that learning is enhanced among high synthesis capacity individuals and by methylphenidate, but impaired by sulpiride. Methylphenidate also blunts implicit effort cost learning. Computational modeling reveals that individuals with higher dopamine synthesis capacity rely more on WM, while methylphenidate boosts their RL rates. The D2 receptor antagonist sulpiride reduces accuracy due to diminished WM involvement and faster WM decay. We conclude that dopamine enhances both slow RL, and fast WM, by promoting plasticity and reducing implicit effort sensitivity. This work was completed as part of a registered trial with the Overview of Medical Research in the Netherlands (NL-OMON43196).

  PublisherOA PDFDOI

• How Working Memory and Reinforcement Learning Interact when Avoiding Punishment and Pursuing Reward Concurrently

  2025-06-02

  preprintOpen accessSenior author

  Humans learn adaptive behaviors via a durable but incremental reinforcement-learning (RL) system and a fast but fleeting working memory (WM) system. Past work parsing these systems focused on reward learning alone, hence little is known about how they interact while simultaneously learning to avoid punishment, and whether arbitrating between these demands is disrupted by psychiatric symptoms. We administered a novel reward/punishment RL-WM task to an online sample oversampled for depression and anxiety symptoms (N=298; n=275 after quality control). Participants avoided punishment during initial learning, yet poorly retained this avoidance. Computational modeling captured this pattern via the fleeting WM system facilitating punishment avoidance, while the durable RL system retained little about punishment. Our task also included two test phases interleaved with learning, which permitted a targeted examination of past findings that WM blunts the RL system. When RL-based retention was tested midway through learning, we indeed found evidence of blunting. Yet, after learning resumed—leading to further prediction errors—blunting was no longer evident in a final test phase. However, individual differences moderated this effect: some individuals were especially susceptible to blunting; for others, WM actually facilitated retention. Finally, task performance was largely spared as a function of depression/anxiety and trait rumination. Overall, our findings demonstrate that—when seeking to attain reward and avoiding punishment concurrently—the WM system can facilitate short-term punishment avoidance while the RL system retains little about punishment; reveal individual differences in the extent to which WM blunts RL; and demonstrate intact behavior under internalizing-disorder symptoms.

  PublisherOA PDFDOI

• eLife Assessment: Neural mechanisms of credit assignment for delayed outcomes during contingent learning

  2025-03-06

  peer-reviewOpen access1st authorCorresponding

  Adaptive behavior in complex environments critically relies on the ability to appropriately link specific choices or actions to their outcomes. However, the neural mechanisms that support the ability to credit only those past choices believed to have caused the observed outcomes remain unclear. Here, we leverage multivariate pattern analyses of functional magnetic resonance imaging (fMRI) data and an adaptive learning task to shed light on the underlying neural mechanisms of such specific credit assignment. We find that the lateral orbitofrontal cortex (lOFC) and hippocampus (HC) code for the causal choice identity when credit needs to be assigned for choices that are separated from outcomes by a long delay, even when this delayed transition is punctuated by interim decisions. Further, we show when interim decisions must be made, learning is additionally supported by lateral frontopolar cortex (lFPC). Our results indicate that lFPC holds previous causal choices in a “pending” state until a relevant outcome is observed, and the fidelity of these representations predicts the fidelity of subsequent causal choice representations in lOFC and HC during credit assignment. Together, these results highlight the importance of the timely reinstatement of specific causes in lOFC and HC in learning choice-outcome relationships when delays and choices intervene, a critical component of real-world learning and decision making.

  PublisherDOI

Recent grants

• Electrophysiological and Computational studies on action monitoring

  NSF · $644k · 2011–2015

• Training Program for Interactionist Cognitive Neuroscience (ICoN)

  NIH · $1.4M · 2019–2029

• How prefrontal cortex augments reinforcement learning

  NSF · $593k · 2015–2019

Frequent coauthors

• Blas Larrauri

  ASM International

  85 shared

• Anne Collins

  University of California, Berkeley

  83 shared

• Samuel Luyasu

  Centre Hospitalier de Luxembourg

  81 shared

• Iris Leibovich‐Nassi

  Barzilai Medical Center

  81 shared

• James F. Cavanagh

  78 shared

• James M. Gold

  70 shared

• Matthew R. Nassar

  Allen Institute for Brain Science

  69 shared

• C. Garren Hester

  67 shared

Awards & honors

• Kavli Fellow (2016)
• Cognitive Neuroscience Society Young Investigator Award (201…
• Janet T Spence Award for early career transformative contrib…
• DG Marquis award for best paper published in Behavioral Neur…