Praveen Suthaharan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Praveen Suthaharan has a total of 54 co-authors. Here are the top co-authors by number of shared publications:
- Philip Robert Corlett
Yale University
15 shared publications - Santiago Castiello
University of Oxford
6 shared publications Joshua Kenney
3 shared publications- Pantelis Leptourgos
Department of Psychiatry, Connecticut Mental Health Center, Yale University, New Haven, CT
3 shared publications Rosa Rossi-Goldthorpe
3 shared publications- Cecilia Heyes
University of Oxford
2 shared publications
Latest Publications
- Open Access
- 30 September 2026
- Published by Center for Open Science in PsyArXiv
Delusion-related aberrant prediction errors are bottom-up
Background: Delusions are the fixed false beliefs that characterize serious mental illnesses like schizophrenia. Prediction errors are surprise signals that guide learning. They have been observed during causal learning in the right dorsolateral prefrontal cortex (rDLPFC). They are registered inappropriately in people with delusions, driving learning towards irrelevant events. This learning correlates with delusion severity. We do not know, however, whether the rDLPFC is receiving the error signal, bottom-up, from other nodes, or, alternatively, whether rDLPFC prediction errors are communicated top-down from rDLPFC. It is also possible that the direction of processing differs between patients with delusions compared with otherwise healthy individuals who espouse delusion-like beliefs but are not delusional per se. Methods: Here, we attempt to distinguish these possibilities using cognitive computational neuroimaging, functional, and effective connectivity analyses of task-based fMRI data in people with schizophrenia (SZ, n=29), people with extreme esoteric beliefs (EEB, n=25), and controls with neither (CTR, n=24), focusing on prediction error processing. Results: During Kamin Blocking of causal learning people who endorse delusions evince aberrant prediction error signals in right dorsolateral prefrontal cortex and anterior insula which correlate with delusion severity. Psychophysiological interaction analysis reveals errant coupling between these regions in people with delusions. Effective connectivity analysis locates the dysfunction in schizophrenia to bottom-up prediction error drive from insula to DLPFC. Conclusion: These data provide quantitative computational metrics for distinguishing delusions from other less problematic beliefs and establish that, in schizophrenia, aberrant prediction errors during causal learning arise from errant bottom-up drive from the insula to the rDLPFC. This may then serve as a marker for clinically relevant odd belief and a target for treatment.
- 30 September 2026
- Published by Center for Open Science in PsyArXiv
Open AccessDelusion-related aberrant prediction errors are bottom-up
Background: Delusions are the fixed false beliefs that characterize serious mental illnesses like schizophrenia. Prediction errors are surprise signals that guide learning. They have been observed during causal learning in the right dorsolateral prefrontal cortex (rDLPFC). They are registered inappropriately in people with delusions, driving learning towards irrelevant events. This learning correlates with delusion severity. We do not know, however, whether the rDLPFC is receiving the error signal, bottom-up, from other nodes, or, alternatively, whether rDLPFC prediction errors are communicated top-down from rDLPFC. It is also possible that the direction of processing differs between patients with delusions compared with otherwise healthy individuals who espouse delusion-like beliefs but are not delusional per se. Methods: Here, we attempt to distinguish these possibilities using cognitive computational neuroimaging, functional, and effective connectivity analyses of task-based fMRI data in people with schizophrenia (SZ, n=29), people with extreme esoteric beliefs (EEB, n=25), and controls with neither (CTR, n=24), focusing on prediction error processing. Results: During Kamin Blocking of causal learning people who endorse delusions evince aberrant prediction error signals in right dorsolateral prefrontal cortex and anterior insula which correlate with delusion severity. Psychophysiological interaction analysis reveals errant coupling between these regions in people with delusions. Effective connectivity analysis locates the dysfunction in schizophrenia to bottom-up prediction error drive from insula to DLPFC. Conclusion: These data provide quantitative computational metrics for distinguishing delusions from other less problematic beliefs and establish that, in schizophrenia, aberrant prediction errors during causal learning arise from errant bottom-up drive from the insula to the rDLPFC. This may then serve as a marker for clinically relevant odd belief and a target for treatment.
- Open Access
- 16 September 2026
- Published by Elsevier in SSRN Electronic Journal
An Interpretable Artificial Intelligence Workflow for Metacognitive Reflection Feedback
As digital learning becomes more interactive, learners are increasingly asked to explain what they understood, where they struggled, and how they approached a task. These open-ended reflections contain valuable traces of metacognition. Advances in large language models (LLMs) and agentic artificial intelligence (AI) now make it possible to score such responses and generate support at scale. However, LLM-scoring often relies on reasoning shortcuts that do not match the processes used by human evaluators. Recent efforts have introduced structured systems that make LLM scoring more transparent and closer to human reasoning. Yet current systems still fall short of providing a clear interpretive workflow that links reflection to actionable, personalized support. We propose a multi-agent metacognitive AI engine that combines probabilistic multidimensional scoring, weakdimension detection, interpretable diagnostic tagging, and scaffold selection with learner-centered personalization. By making each stage evidence-linked and inspectable, the workflow connects reflective language to pedagogical action without collapsing assessment, diagnosis, and feedback into a single black box decision. This framework offers a testable design agenda for improving construct validity, interpretability, and actionable support in AI-mediated learning environments.
- 16 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAn Interpretable Artificial Intelligence Workflow for Metacognitive Reflection Feedback
As digital learning becomes more interactive, learners are increasingly asked to explain what they understood, where they struggled, and how they approached a task. These open-ended reflections contain valuable traces of metacognition. Advances in large language models (LLMs) and agentic artificial intelligence (AI) now make it possible to score such responses and generate support at scale. However, LLM-scoring often relies on reasoning shortcuts that do not match the processes used by human evaluators. Recent efforts have introduced structured systems that make LLM scoring more transparent and closer to human reasoning. Yet current systems still fall short of providing a clear interpretive workflow that links reflection to actionable, personalized support. We propose a multi-agent metacognitive AI engine that combines probabilistic multidimensional scoring, weakdimension detection, interpretable diagnostic tagging, and scaffold selection with learner-centered personalization. By making each stage evidence-linked and inspectable, the workflow connects reflective language to pedagogical action without collapsing assessment, diagnosis, and feedback into a single black box decision. This framework offers a testable design agenda for improving construct validity, interpretability, and actionable support in AI-mediated learning environments.
- Open Access
- 10 February 2026
- Published by Center for Open Science in PsyArXiv
Exaggerated Volatility Beliefs Drive Social Hallucinations in Paranoia
The social content of paranoid delusions (that others are ‘out to get me’) has led some to propose that paranoia stems from specific abnormalities in social cognition. For instance, in ‘social hallucinations’ paranoid perceivers are more likely to falsely attribute harmful intention and agency to unstructured, randomly moving visual patterns. However, recent work has revealed that paranoia is marked by domain-general changes in belief updating that affect learning and inference across all contexts. Whether domain-general changes in paranoid cognition are sufficient to explain social hallucinations—without invoking breakdowns in specifically-social cognitive processes—has not been systematically investigated. To test this possibility, across two experiments we examined probabilistic reversal learning and perceived animacy in chasing displays in individuals with high and low levels of paranoia. Participants completed one of four probabilistic reversal learning pre-training conditions (social/non-social × easy/hard) before a perceived animacy task where high confidence false alarms indicated an individual’s proneness to social hallucinations. High-paranoia participants showed increased win-switching in probabilistic reversal learning and elevated high-confidence false alarms in perceived animacy, with atypicalities in general learning behaviour predicting the severity of social hallucinations —suggesting linked mechanisms. Hierarchical Gaussian Filter modeling revealed context-dependent mechanisms: when volatility was salient, prior volatility beliefs (μ₃) linked win-switching, social hallucinations, and paranoia; when social context was salient, meta-volatility learning rates (ω₃) mediated these relationships. These findings are consistent with a picture where social hallucinations have their genesis in domain-general belief updating deficits, advancing our mechanistic understanding of how paranoid percepts and cognitions arise.
- 10 February 2026
- Published by Center for Open Science in PsyArXiv
Open AccessExaggerated Volatility Beliefs Drive Social Hallucinations in Paranoia
The social content of paranoid delusions (that others are ‘out to get me’) has led some to propose that paranoia stems from specific abnormalities in social cognition. For instance, in ‘social hallucinations’ paranoid perceivers are more likely to falsely attribute harmful intention and agency to unstructured, randomly moving visual patterns. However, recent work has revealed that paranoia is marked by domain-general changes in belief updating that affect learning and inference across all contexts. Whether domain-general changes in paranoid cognition are sufficient to explain social hallucinations—without invoking breakdowns in specifically-social cognitive processes—has not been systematically investigated. To test this possibility, across two experiments we examined probabilistic reversal learning and perceived animacy in chasing displays in individuals with high and low levels of paranoia. Participants completed one of four probabilistic reversal learning pre-training conditions (social/non-social × easy/hard) before a perceived animacy task where high confidence false alarms indicated an individual’s proneness to social hallucinations. High-paranoia participants showed increased win-switching in probabilistic reversal learning and elevated high-confidence false alarms in perceived animacy, with atypicalities in general learning behaviour predicting the severity of social hallucinations —suggesting linked mechanisms. Hierarchical Gaussian Filter modeling revealed context-dependent mechanisms: when volatility was salient, prior volatility beliefs (μ₃) linked win-switching, social hallucinations, and paranoia; when social context was salient, meta-volatility learning rates (ω₃) mediated these relationships. These findings are consistent with a picture where social hallucinations have their genesis in domain-general belief updating deficits, advancing our mechanistic understanding of how paranoid percepts and cognitions arise.
- Open Access
- 16 January 2026
- Published by Ubiquity Press, Ltd. in Computational Psychiatry
- Vol. 10 (1) , 104-120
Excessive switching in OCD and paranoia arises from different deficits in belief-updating
Both obsessions and paranoia are characterized by cognitive inflexibility, particularly in uncertain environments. Yet, differential diagnosis is challenging and limited to clinical interviews and self-report symptom questionnaires. We predicted that obsessions and paranoia would be associated with distinct patterns of behavior in our well-established probabilistic reversal learning (PRL) task. Probabilistic reversal learning involves updating beliefs about rewards when contingencies change. Obsessions and paranoia have been linked to excessive switching behaviors during reversal learning, although some report perseveration in patients with OCD and those with schizophrenia. Here, we analyze data gathered from the general population to assess the associations between obsessions, paranoia, and PRL task performance. Using a novel computational method – Bayesian Gaussian Graphical Modeling combined with a Hierarchical Gaussian Filter – we distinguish the impacts of paranoia and obsession on reversal learning – despite their significant correlation. We find that win-switching arises in paranoia from deficits in learning about uncertainty in the global task structure, whereas excessive switching in OCD arises from challenges in learning about uncertainty at the local choice level.
- 16 January 2026
- Published by Ubiquity Press, Ltd. in Computational Psychiatry
- Vol. 10 (1) , 104-120
Open AccessExcessive switching in OCD and paranoia arises from different deficits in belief-updating
Both obsessions and paranoia are characterized by cognitive inflexibility, particularly in uncertain environments. Yet, differential diagnosis is challenging and limited to clinical interviews and self-report symptom questionnaires. We predicted that obsessions and paranoia would be associated with distinct patterns of behavior in our well-established probabilistic reversal learning (PRL) task. Probabilistic reversal learning involves updating beliefs about rewards when contingencies change. Obsessions and paranoia have been linked to excessive switching behaviors during reversal learning, although some report perseveration in patients with OCD and those with schizophrenia. Here, we analyze data gathered from the general population to assess the associations between obsessions, paranoia, and PRL task performance. Using a novel computational method – Bayesian Gaussian Graphical Modeling combined with a Hierarchical Gaussian Filter – we distinguish the impacts of paranoia and obsession on reversal learning – despite their significant correlation. We find that win-switching arises in paranoia from deficits in learning about uncertainty in the global task structure, whereas excessive switching in OCD arises from challenges in learning about uncertainty at the local choice level.
- Open Access
- 16 January 2026
- Published by Ubiquity Press, Ltd. in Computational Psychiatry
- Vol. 10 (1) , 18-34
Self-Reflection Protects Behavior from Volatile Beliefs Linked to Paranoia
Processing uncertainty may be pathognomonic (characteristic of a disease) for some psychiatric conditions. Some people expect the world to change, even when it doesn’t. This tendency is central to paranoia, where individuals often anticipate threat or change without clear evidence. But what determines whether these beliefs translate into behavior? One possibility is that metacognitive structure – the coherence and depth with which one articulates their own thinking – acts as a buffer. An agent may endorse a belief but have sufficient accessory hypotheses to insulate it from action. To test this, we used metacognitive prompting in GPT-4 to score individual reflections on open-ended questions (e.g., did you use any particular strategy?) after completing a probabilistic reversal learning task. Individuals with higher paranoia demonstrate lower metacognitive structure (t = 5.98, p < 0.001), with metacognition attenuating the relationship between volatility belief and switching behavior (Δ = –15 pp, p < 0.001) even after controlling for reflection verbosity and general cognitive ability. These findings suggest that metacognition protects against uncertainty-driven instability, pointing to a key mechanism by which reflection protects against cognition under change. This work provides a novel framework to measure metacognition from behavioral task debrief questions.
- 16 January 2026
- Published by Ubiquity Press, Ltd. in Computational Psychiatry
- Vol. 10 (1) , 18-34
Open AccessSelf-Reflection Protects Behavior from Volatile Beliefs Linked to Paranoia
Processing uncertainty may be pathognomonic (characteristic of a disease) for some psychiatric conditions. Some people expect the world to change, even when it doesn’t. This tendency is central to paranoia, where individuals often anticipate threat or change without clear evidence. But what determines whether these beliefs translate into behavior? One possibility is that metacognitive structure – the coherence and depth with which one articulates their own thinking – acts as a buffer. An agent may endorse a belief but have sufficient accessory hypotheses to insulate it from action. To test this, we used metacognitive prompting in GPT-4 to score individual reflections on open-ended questions (e.g., did you use any particular strategy?) after completing a probabilistic reversal learning task. Individuals with higher paranoia demonstrate lower metacognitive structure (t = 5.98, p < 0.001), with metacognition attenuating the relationship between volatility belief and switching behavior (Δ = –15 pp, p < 0.001) even after controlling for reflection verbosity and general cognitive ability. These findings suggest that metacognition protects against uncertainty-driven instability, pointing to a key mechanism by which reflection protects against cognition under change. This work provides a novel framework to measure metacognition from behavioral task debrief questions.
- Open Access
- 11 September 2025
- Published by Center for Open Science in PsyArXiv
Pseudo-specificity, pseudo-modules and pseudo-models in paranoia
A reply to Razavi et al: Pseudo-approaches lead to pseudo-explanations
- 11 September 2025
- Published by Center for Open Science in PsyArXiv
Open AccessPseudo-specificity, pseudo-modules and pseudo-models in paranoia
A reply to Razavi et al: Pseudo-approaches lead to pseudo-explanations
- Open Access
- 4 July 2025
- Published by Center for Open Science in PsyArXiv
Self-reflection protects behavior from volatile beliefs linked to paranoia
Processing uncertainty may be pathognomonic (characteristic of a disease) for some psychiatric conditions. Some people expect the world to change, even when it doesn’t. This tendency is central to paranoia, where individuals often anticipate threat or change without clear evidence. But what determines whether these beliefs translate into behavior? One possibility is that metacognitive structure – the coherence and depth with which one articulates their own thinking – acts as a buffer. An agent may endorse a belief but have sufficient accessory hypotheses to insulate it from action. To test this, we used metacognitive prompting in GPT-4 to score individual reflections on open-ended questions (e.g., did you use any particular strategy?) after completing a probabilistic reversal learning task. Individuals with higher paranoia demonstrate lower metacognitive structure (t = 5.98, p < 0.001), with metacognition attenuating the relationship between volatility belief and switching behavior (Δ = -15 pp, p < 0.001) even after controlling for reflection verbosity and general cognitive ability. These findings suggest that metacognition protects against uncertainty-driven instability, pointing to a key mechanism by which reflection protects against cognition under change. This work provides a novel framework to measure metacognition from behavioral task debrief questions.
- 4 July 2025
- Published by Center for Open Science in PsyArXiv
Open AccessSelf-reflection protects behavior from volatile beliefs linked to paranoia
Processing uncertainty may be pathognomonic (characteristic of a disease) for some psychiatric conditions. Some people expect the world to change, even when it doesn’t. This tendency is central to paranoia, where individuals often anticipate threat or change without clear evidence. But what determines whether these beliefs translate into behavior? One possibility is that metacognitive structure – the coherence and depth with which one articulates their own thinking – acts as a buffer. An agent may endorse a belief but have sufficient accessory hypotheses to insulate it from action. To test this, we used metacognitive prompting in GPT-4 to score individual reflections on open-ended questions (e.g., did you use any particular strategy?) after completing a probabilistic reversal learning task. Individuals with higher paranoia demonstrate lower metacognitive structure (t = 5.98, p < 0.001), with metacognition attenuating the relationship between volatility belief and switching behavior (Δ = -15 pp, p < 0.001) even after controlling for reflection verbosity and general cognitive ability. These findings suggest that metacognition protects against uncertainty-driven instability, pointing to a key mechanism by which reflection protects against cognition under change. This work provides a novel framework to measure metacognition from behavioral task debrief questions.
- Open Access
- 13 February 2025
- Published by Center for Open Science in PsyArXiv
Parasocial Cognition and Paranoia
It has been argued that social processes are relevant to belief formation and maintenance and thence to persecutory delusions – the fixed false beliefs that others intend harm. We call this the social turn in delusions research. It suggests that paranoia is the purview of a specialized mechanism for coalitional cognition – thinking about group membership and reputation management. Here we suggest instead that a simpler, parasocial, learning mechanism may underwrite persecutory and other delusions. We make our case in terms of computations (prediction, not coalition), algorithm (association rather than recursion), and implementation (dopaminergic domain-general rather than social-specific regions). We conclude with suggestions for adversarial collaboration that will clarify the contributions of domain-general versus social-specific processes to delusions.
- 13 February 2025
- Published by Center for Open Science in PsyArXiv
Open AccessParasocial Cognition and Paranoia
It has been argued that social processes are relevant to belief formation and maintenance and thence to persecutory delusions – the fixed false beliefs that others intend harm. We call this the social turn in delusions research. It suggests that paranoia is the purview of a specialized mechanism for coalitional cognition – thinking about group membership and reputation management. Here we suggest instead that a simpler, parasocial, learning mechanism may underwrite persecutory and other delusions. We make our case in terms of computations (prediction, not coalition), algorithm (association rather than recursion), and implementation (dopaminergic domain-general rather than social-specific regions). We conclude with suggestions for adversarial collaboration that will clarify the contributions of domain-general versus social-specific processes to delusions.
- Open Access
- 10 October 2023
- Published by Center for Open Science in PsyArXiv
Lesions to the mediodorsal thalamus but not orbitofrontal cortex enhance volatility beliefs linked to paranoia
Beliefs – attitudes toward some state of the environment – guide action selection and should be robust to variability but sensitive to meaningful change. Beliefs about volatility (expectation of change) are associated with paranoia in humans yet the brain regions responsible for volatility beliefs remain unknown. Orbitofrontal cortex (OFC) is central to adaptive behavior whereas magnocellular mediodorsal thalamus (MDmc) is essential for arbitrating between perceptions and action policies. We assessed belief updating in a three-choice probabilistic reversal-learning task following excitotoxic lesions of MDmc (n=3) or OFC (n=3) and compared performance with that of unoperated rhesus macaques (n=14). Computational analyses indicated that lesions of the MDmc, but not OFC, were associated with erratic switching behavior and heightened volatility belief (as in paranoia in humans). In contrast, OFC lesions were associated with increased lose-stay behavior and reward learning rates. Given the consilience across species and models these results have implications for understanding paranoia.
- 10 October 2023
- Published by Center for Open Science in PsyArXiv
Open AccessLesions to the mediodorsal thalamus but not orbitofrontal cortex enhance volatility beliefs linked to paranoia
Beliefs – attitudes toward some state of the environment – guide action selection and should be robust to variability but sensitive to meaningful change. Beliefs about volatility (expectation of change) are associated with paranoia in humans yet the brain regions responsible for volatility beliefs remain unknown. Orbitofrontal cortex (OFC) is central to adaptive behavior whereas magnocellular mediodorsal thalamus (MDmc) is essential for arbitrating between perceptions and action policies. We assessed belief updating in a three-choice probabilistic reversal-learning task following excitotoxic lesions of MDmc (n=3) or OFC (n=3) and compared performance with that of unoperated rhesus macaques (n=14). Computational analyses indicated that lesions of the MDmc, but not OFC, were associated with erratic switching behavior and heightened volatility belief (as in paranoia in humans). In contrast, OFC lesions were associated with increased lose-stay behavior and reward learning rates. Given the consilience across species and models these results have implications for understanding paranoia.
- Open Access
- 14 September 2022
- Published by Center for Open Science in PsyArXiv
Reasoning biases, behavior, and computation in delusions: shared and unique variance
Introduction: Multiple measures of decision-making under uncertainty (e.g. jumping to conclusions, bias against disconfirmatory evidence, win-switch behavior, random exploration) have been associated with delusional thinking in independent studies. Yet, it is unknown whether these variables explain shared or unique variance in delusional thinking, and whether these relationships are specific to paranoia or unusual thought content more broadly. Additionally, the underlying computational mechanisms require further investigation.Methods: Decision-making data were collected in 88 individuals (46 healthy controls, 42 schizophrenia-spectrum) and included measures of reasoning biases (self-reported cognitive insight and jumping to conclusions (JTC) and task-based bias against disconfirmatory evidence (BADE)) and behavior on probabilistic reversal learning (PRL) and explore/exploit tasks. Stepwise regression was used to determine which variables were most strongly, uniquely associated with paranoia and unusual thought content in schizophrenia patients alone, and across the whole sample. Results: Win-switch rate during PRL was the only cognitive measure that was significantly different between groups. Reversal learning performance, random exploration, and poor evidence integration during BADE showed significant, independent associations with paranoia. Only JTC was associated with unusual thought content. Computational parameters increased the proportion of variance explained in paranoia. Conclusions: Decision-making influenced by strong volatility and variability is specifically associated with paranoia, whereas hasty decision-making is specifically associated with unusual thought content. These aspects of decision-making under uncertainty may therefore represent distinct cognitive processes that, together, have the potential to worsen delusional thinking across the psychosis spectrum.
- 14 September 2022
- Published by Center for Open Science in PsyArXiv
Open AccessReasoning biases, behavior, and computation in delusions: shared and unique variance
Introduction: Multiple measures of decision-making under uncertainty (e.g. jumping to conclusions, bias against disconfirmatory evidence, win-switch behavior, random exploration) have been associated with delusional thinking in independent studies. Yet, it is unknown whether these variables explain shared or unique variance in delusional thinking, and whether these relationships are specific to paranoia or unusual thought content more broadly. Additionally, the underlying computational mechanisms require further investigation.Methods: Decision-making data were collected in 88 individuals (46 healthy controls, 42 schizophrenia-spectrum) and included measures of reasoning biases (self-reported cognitive insight and jumping to conclusions (JTC) and task-based bias against disconfirmatory evidence (BADE)) and behavior on probabilistic reversal learning (PRL) and explore/exploit tasks. Stepwise regression was used to determine which variables were most strongly, uniquely associated with paranoia and unusual thought content in schizophrenia patients alone, and across the whole sample. Results: Win-switch rate during PRL was the only cognitive measure that was significantly different between groups. Reversal learning performance, random exploration, and poor evidence integration during BADE showed significant, independent associations with paranoia. Only JTC was associated with unusual thought content. Computational parameters increased the proportion of variance explained in paranoia. Conclusions: Decision-making influenced by strong volatility and variability is specifically associated with paranoia, whereas hasty decision-making is specifically associated with unusual thought content. These aspects of decision-making under uncertainty may therefore represent distinct cognitive processes that, together, have the potential to worsen delusional thinking across the psychosis spectrum.