Scroll To Top
Scroll To Top

Computational Psychiatry: New Perspectives on Mental Illness

edited by A. David Redish and Joshua A. Gordon

Published by MIT Press Hardcover ISBN:9780262035422 eBook  ISBN: 9780262337847

1 - On the Cusp: Current Challenges and Promises in Psychiatry - PDF

Joshua A. Gordon and A. David Redish

2 - Breakdowns and Failure Modes: An Engineer’s View - PDF

A. David Redish and Joshua A. Gordon

Open Issues in Psychiatry

3 - Complexity and Heterogeneity in Psychiatric Disorders: Opportunities for Computational Psychiatry - Abstract - PDF

Nelson Totah, Huda Akil, Quentin J. M. Huys, John H. Krystal, Angus W. MacDonald III, Tiago V. Maia, Robert C. Malenka, and Wolfgang M. Pauli

4 - What Does Computational Psychiatry Need to Explain to Capture Mechanisms of Psychopathology? Facts, Almost Facts, and Hints - Abstract

Deanna M. Barch


5 - Computational Approaches for Studying Mechanisms of Psychiatric Disorders - Abstract - PDF

Zeb Kurth-Nelson, John P. O’Doherty, Deanna M. Barch, Sophie Denève, Daniel Durstewitz, Michael J. Frank, Joshua A. Gordon, Sanjay J. Mathew, Yael Niv, Kerry Ressler, and Heike Tost

6 - Computational Cognitive Neuroscience Approaches to Deconstructing Mental Function and Dysfunction - Abstract - PDF

Michael J. Frank

7 - How Could We Get Nosology from Computation? - Abstract

Christoph Mathys


8 - Current State of Psychiatric Nosology - Abstract

Michael B. First

9 - The Computation of Collapse: Can Reliability Engineering Shed Light on Mental Illness? - Abstract

Angus W. MacDonald III, Jennifer L. Zick, Theoden I. Netoff, and Matthew V. Chafee

10 - A Novel Framework for Improving Psychiatric Diagnostic Nosology - Abstract - PDF

Shelly B. Flagel, Daniel S. Pine, Susanne E. Ahmari, Michael B. First, Karl J. Friston, Christoph Mathys, A. David Redish, Katharina Schmack, Jordan W. Smoller, and Anita Thapar

11 - Computational Nosology and Precision Psychiatry: A Proof of Concept - Abstract - PDF

Karl J. Friston


12 - Candidate Examples for a Computational Approach to Address Practical Problems in Psychiatry - Abstract

Rosalyn Moran, Klaas Enno Stephan, Matthew Botvinick, Michael Breakspear, Cameron S. Carter, Peter W. Kalivas, P. Read Montague, Martin P. Paulus, and Frederike Petzschner

13 - There Are No Killer Apps but Connecting Neural Activity to Behavior through Computation Is Still a Good Idea - Abstract

P. Read Montague

14 - Call for Pragmatic Computational Psychiatry: Integrating Computational Approaches and Risk-Prediction Models
and Disposing of Causality - Abstract - PDF

Martin P. Paulus, Crane Huang, and Katia M. Harlé

15 - A Valuation Framework for Emotions Applied to Depression and Recurrence - Abstract

Quentin J. M. Huys

16 - Clinical Heterogeneity Arising from Categorical and Dimensional Features of the Neurobiology of Psychiatric Diagnoses: Insights from Neuroimaging and Computational Neuroscience - Abstract - PDF

John H. Krystal, Alan Anticevic, John D. Murray, David Glahn, Naomi Driesen, Genevieve Yang, and Xiao-Jing Wang


17 - From Psychiatry to Computation and Back Again - PDF

A. David Redish and Joshua A. Gordon

18 - Bibliography - PDF

Psychiatry faces a number of challenges, among them are the reconceptualization of symptoms and diagnoses, disease prevention, treatment development and monitoring of its effects, and the provision of individualized, precision medicine. Achieving these goals will require an increase in the biological, quantitative, and theoretical grounding of psychiatry. To address these challenges, psychiatry must confront the complexity and heterogeneity intrinsic to the nature of brain disorders. This chapter seeks to identify the sources of complexity and heterogeneity as a means of confronting the challenges facing the field. These sources include the interplay between genetic and epigenetic factors with the environment and their impact on neural circuits. Moreover, these interactions are expressed dynamically over the course of development and continue to play out during the disease process and treatment.
We propose that computational approaches provide a framework for addressing the complexity and heterogeneity that underlie the challenges facing psychiatry. Central to our argument is the idea that these characteristics are not noise to be eliminated from diagnosis and treatment of disorders. Instead, such complexity and heterogeneity arises from intrinsic features of brain function and, therefore, represent opportunities for computational models to provide a more accurate biological foundation for diagnosis and treatment of psychiatric disorders. The challenges to be addressed by a computational framework include the following. First, it must improve the search for risk factors and biomarkers, which can be used toward primary prevention of disease. Second, it must help to represent the biological ground truth of psychiatric disorders, which will improve the accuracy of diagnostic categories, assist in discovering new treatments, and aid in precision medicine. Third, to be useful for secondary prevention, it must represent how risk factors, biomarkers, and the underlying biology change through the course of development, disease progression, and treatment process.
This chapter provides specific research examples on the neurobiology of mental illness—using psychosis as a case in point—that may begin to rise to the level of “facts,” or at least “almost facts” or strong “hints,” about important etiological mechanisms that need to be explained to capture key components of at least some facets of mental illness. These examples are then used to illustrate where computational psychiatry approaches may help. In particular, there is an opportunity to provide links across different levels of analysis (e.g., behavior, systems level, specific circuits and even genetic influences) in ways that can lead to a more unified framework for understanding the apparent multitude of impairments present in psychosis, which may in turn lead to the identification of new treatment or even prevention targets. This chapter also discusses some of the known conundrums about the etiology of mental illness that need to be accounted for in computational frameworks, including the presence of heterogeneity within current diagnostic categories, the vast degree of comorbidity across current diagnostic categories, and the need to reconceptualize the dimensionality versus categorical nature of mental illness.
Vast spectra of biological and psychological processes are potentially involved in the mechanisms of psychiatric illness. Computational neuroscience brings a diverse toolkit to bear on understanding these processes. This chapter begins by organizing the many ways in which computational neuroscience may provide insight to the mechanisms of psychiatric illness. It then contextualizes the quest for deep mechanistic understanding through the perspective that even partial or nonmechanistic understanding can be applied productively. Finally, it questions the standards by which these approaches should be evaluated. If computational psychiatry hopes to go beyond traditional psychiatry, it cannot be judged solely on the basis of how closely it reproduces the diagnoses and prognoses of traditional psychiatry, but must also be judged against more fundamental measures such as patient outcomes.
Advances in our understanding of brain function and dysfunction require the integration of heterogeneous sources of data across multiple levels of analysis, from biophysics to cognition and back. This chapter reviews the utility of computational neuroscience approaches across these levels and how they have advanced our understanding of multiple constructs relevant for mental illness, including working memory, reward-based decision making, model-free and model-based reinforcement learning, exploration versus exploitation, Pavlovian contributions to motivated behavior, inhibitory control, and social interactions. The computational framework formalizes these processes, providing quantitative and falsifiable predictions. It also affords a characterization of mental illnesses not in terms of overall deficit but rather in terms of aberrations in managing fundamental trade-offs inherent within healthy cognitive processing.
Psychiatry has found it difficult to develop a nosology that allows for the targeted treatment of disorders of the mind. The historic inability of the field to agree on a nosology based on clinical experience has led it to retreat to diagnoses based on symptom checklists as laid down in the Diagnostic and Statistical Manual of Mental Disorders (DSM). While this has increased the reliability of diagnoses, hopes that biological findings would lead to the emergence of mechanistically founded diagnostic entities have not been realized despite considerable advances in neurobiology. This article sets out a possible way forward: harnessing systems theory to provide the conceptual constraints needed to link clinical phenomena with neurobiology. This approach builds on the insight that the mind is a system which, to regulate its environment, needs to have a model of that environment and needs to update predictions about it using the rules of inductive logic (i.e., Bayesian inference). The application of the rules of inductive logic is called Bayesian inference because Bayes’s theorem is the most important consequence of these rules, prescribing how beliefs need to be updated in response to new information. Importantly, while Bayesian inference is by definition consistent with the rules of inductive logic, it can still be false (to the point of being pathological), in the sense of leading to false predictions, because the model underlying the inference is inadequate. Further, it can be shown that Bayesian inference can be reduced to updating beliefs based on precision-weighted prediction errors, where a prediction error is the difference between actual and predicted input, and precision is the confidence associated with the input prediction. Precision weighting of prediction errors entails that a given discrepancy between outcome and prediction means more, and leads to greater belief updates, the more confidently the prediction was made. This provides a conceptual framework linking clinical experience with the pathophysiology underlying disorders of the mind. Limitations of this approach are discussed and ways to work around them illustrated with examples. Finally, initial steps and possible future directions toward a nosology based on failures of precision weighting are discussed.
Psychiatric classifications categorize how patients present to mental healthcare professionals and are necessarily utilitarian. From the clinician’s perspective, the most important goal of a psychiatric classification is to assist them in managing their patients’ psychiatric conditions by facilitating the selection of effective interventions and predicting management needs and outcomes. Due to the field’s lack of understanding of the neurobiological mechanisms underlying the psychiatric disorders in both the Diagnostic and Statistical Manual of Mental Disorders (DSM) and the International Classification of Diseases (ICD), diagnosis and treatment are only loosely related, thus limiting clinical utility. Both DSM and the chapter on mental and behavioral disorders in ICD adopted a descriptive atheoretical categorical approach that defines mental disorders according to syndromal patterns of presenting symptoms. This chapter discusses the fundamental challenges that underlie this decision. It then reviews the Research Domain Criteria (RDoC) project, a research framework established by the U.S. National Institute of Mental Health (NIMH) to assist researchers in relating the fundamental domains of behavioral functioning to their underlying neurobiological components. Designed to support the acquisition of knowledge of causal mechanisms underlying mental disorders, RDoC may facilitate a future paradigm shift in the classification of mental disorder.
Computational modeling in psychiatry has generally followed from efforts to understand cognitive processes (McClelland and Rumelhart 1986) or the nervous system (Hodgkin and Huxley 1952). This stands to reason: psychiatric disorders are disorders of thought and central nervous system activity. Although there are few contributions to psychiatry from probability theorists and engineers (Shewhart 1938; Miner 1945; Lusser 1958), the tools developed for quality control of metal fatigue and failed rockets may point to a useful approach for thinking about mental illness. This chapter argues that the computational science of collapse, which describes the manner and likelihood of failures in complex systems, provides a framework in which to use computational modeling for relating mechanisms to behavioral outcomes. This science, known as reliability engineering, is a branch of applied probability theory that has now been used for almost a century to help understand and predict how inorganic, complex systems break down. The idea of a fault tree analysis is introduced, a tool developed in reliability engineering which may be able to incorporate and provide a broader structure for more traditional computational models. Finally, Some of the current challenges of psychiatric classification are unpacked, and discussion follows on how this framework might be adapted to provide a unifying framework for classification and etiology.
This chapter proposes a new framework for diagnostic nosology based on Bayesian principles. This novel integrative framework builds upon and improves the current diagnostic system in psychiatry. Instead of starting from the assumption that a diagnosis describes a specific unitary dysfunction that causes a set of symptoms, it is assumed that the underlying disease causes the clinician to make a diagnosis. Thus, unlike the current diagnostic system, this framework treats both symptoms and diagnostic classification as consequences of the underlying pathophysiology. Comorbidities are therefore easily incorporated into the framework and inform, rather than hinder, the diagnostic process. Further, the proposed framework provides a bridge—which did not previously exist—that links putative constructs related to pathophysiology (e.g., RDoC domains) and clinical diagnoses (e.g., DSM categories) related to signs and symptoms. The model is flexible; it is expandable and collapsible, and can integrate a diverse array of data at multiple levels. Crucially, this novel framework explicitly provides an iterative approach, updating and selecting the best model, based on the highest-quality available evidence at any point. In fact, the scheme can, in principle, automatically ignore data that is not relevant or informative to the diagnostic trajectory. Finally, the proposed framework can account for and incorporate the longitudinal course of an illness. This chapter details the theoretical basis for this framework and provides clinical examples to illustrate its utility and application. Multiple iterations of this framework will be required based on available information. It is hoped that, with time, the framework will enhance our understanding of individual differences in brain function and behavior and ultimately improve treatment outcomes in psychiatry.
This chapter provides an illustrative treatment of psychiatric morbidity that offers an alternative to the standard nosological model in psychiatry. It considers what would happen if we treated diagnostic categories not as putative causes of signs and symptoms, but as diagnostic consequences of psychopathology and pathophysiology. This reconstitution (of the standard model) opens the door to a more natural formulation of how patients present and their likely response to therapeutic interventions. The chapter describes a model that generates symptoms, signs, and diagnostic outcomes from latent psychopathological states. In turn, psychopathology is caused by pathophysiological processes that are perturbed by (etiological) causes, such as predisposing factors, life events, and therapeutic interventions. The key advantages of this nosological formulation include: (a) the formal integration of diagnostic (e.g., DSM) categories and latent psychopathological constructs (e.g., the dimensions of RDoC); (b) the provision of a hypothesis or model space that accommodates formal evidence-based hypothesis testing or model selection (using Bayesian model comparison); (c) the ability to predict therapeutic responses (using a posterior predictive density), as in precision medicine; and (d) a framework that allows one to test hypotheses about the interactions between pharmacological and psychotherapeutic interventions. These and other advantages are largely promissory at present: the purpose of this chapter is to show what might be possible, through the use of idealized simulations. These simulations can be regarded as a (conceptual) prospectus that motivates a computational nosology for psychiatry.
Scientists and clinicians can utilize a model-based framework to develop computational approaches to psychiatric practice and bring scientific discoveries to a clinical interface. This chapter describes a general modeling perspective, which complements those derived in previous chapters, and provides distinct examples to highlight the scientific and preclinical research that can evolve out of a computational framework to offer new tools for clinical practice. It begins by reviewing areas of theoretical and modeling studies that have reached a critical mass and outlines the pathophysiological insights that have been revealed. Three particular models are used to demonstrate how clinical questions, relating to understanding disease mechanisms and predicting treatment response, could be potentially addressed using an integrated computational framework. First, the phasic dopamine temporal difference model shows how neurophysiological and neuroanatomical research, incorporated into a learning circuit model, provides a constrained hypothesis testing framework, related to the likely multiple mechanisms contributing to addiction. Second, a potential application of generative models of neuroimaging measurements (dynamic causal models of EEG data) is described to predict individual treatment responses in patients with schizophrenia. The third example offers a novel approach to quantifying patient outcomes under a “recovery model” of psychiatric illness. This involves a dynamical system appraisal of allostasis, using the amygdala-HPA axis with its role in anxiety disorders and depression as a clinical target syndrome to which the model could be applied. In conclusion, consideration is given to the community efforts needed to support the validation of these and future applications.
The quest to understand the relationship between neural activity and behavior has been ongoing for well over a hundred years. Although research based on the stimulus-and-response approach to behavior, advocated by behaviorists, flourished during the last century, this view does not, by design, account for unobservable variables (e.g., mental states). Putting aside this approach, modern cognitive science, cognitive neuroscience, neuroeconomics, and behavioral economics have sought to explain this connection computationally. One major hurdle lies in the fact that we lack even a simple model of cognitive function. This chapter sketches an application that connects neuromodulator function to decision making and the valuation that underlies it. The nature of this hypothesized connection offers a fruitful platform to understand some of the informational aspects of dopamine function in the brain and how it exposes many different ways of understanding motivated choice.
Biological psychiatry is at an impasse. Despite several decades of intense research, few if any, biological parameters have contributed to a significant improvement in the life of a psychiatric patient. It is argued that this impasse may be a consequence of an obsessive focus on mechanisms. Alternatively, a risk-prediction framework provides a more pragmatic approach, because it aims to develop tests and measures which generate clinically useful information. Computational approaches may have an important role to play here. This chapter presents an example of a risk-prediction framework, which shows that computational approaches provide a significant predictive advantage. Future directions and challenges are highlighted.
The burden of depression is substantially aggravated by relapses and recurrences, and these become more inevitable with every episode of depression. This chapter describes how computational psychiatry can provide a normative framework for emotions and an integrative approach to core cognitive components of depression and relapse. Central to this is the notion that emotions effectively imply a valuation; thus they are amenable to description and dissection by reinforcement-learning methods. It is argued that cognitive accounts of emotion can be viewed in terms of model-based valuation, and that automatic emotional responses relate to model-free valuation and the innate recruitment of fixed behavioral patterns. This model-based view captures phenomena such as helplessness, hopelessness, attributions, and stress sensitization. Considering it in more atomic algorithmic detail opens up the possibility of viewing rumination and emotion regulation in this same normative framework. The problem of treatment selection for relapse and recurrence prevention is outlined and suggestions made on how the computational framework of emotions might help improve this. The chapter closes with a brief overview of what we can hope to gain from computational psychiatry.
Clinical heterogeneity presents important challenges to optimizing psychiatric diagnoses and treatments. Patients clustered within current diagnostic schema vary widely on many features of their illness, including their responses to treatments. As outlined by the American Psychiatric Association Diagnostic and Statistical Manual (DSM), psychiatric diagnoses have been refined since DSM was introduced in 1952. These diagnoses serve as the targets for current treatments and supported the emergence of psychiatric genomics. However, the Research Domain Criteria highlight DSM’s shortcomings, including its limited ability to encompass dimensional features linking patients across diagnoses. This chapter considers elements of the dimensional and categorical features of psychiatric diagnoses, with a particular focus on schizophrenia. It highlights ways that computational neuroscience approaches have shed light on both dimensional and categorical features of the biology of schizophrenia. It also considers opportunities and challenges associated with attempts to reduce clinical heterogeneity through categorical and dimensional approaches to clustering patients. Finally, discussion will consider ways that one might work with both approaches in parallel or sequentially, as well as diagnostic schema that might integrate both perspectives.