Philosophy of psychiatry can gain greater vitality by engaging the implications of new science

Stein et al1 introduce philosophy of psychiatry by presenting a series of well-curated, historically grounded positions on topics with which the discipline has been centrally engaged. These include the definition and boundaries of psychopathology, the unsettled nature of diagnostic constructs in psychiatry (including the dialectic between naturalism and normativism), the perils of essentialism, and mind-brain relationships. They introduce these topics partly through the compelling device of asynchronous "conversational" threads containing positions and arguments from successive writers, and accompany this exposition with their own gently reflective comments. They also provide intellectual frameworks to orient readers who are coming to these topics for the first time. For example, they offer a context within which to understand the debate between naturalist and normativist positions on the definitions of psychiatric disorders, a debate that might be less familiar to readers coming from scientifically more mature fields of medicine that confidently ground diagnoses on knowledge of etiology and pathophysiology2. Stein et al compare strong versions of naturalism in psychiatry, which hold that disorders can be defined in purely factual, value-free scientific terms, with strong versions of normativism, which hold that disorder definitions always require value judgments, and which in their most extreme form devolve into the relativism exploited by the antipsychiatry movement of the mid-late 20th century. They then discuss compromise or bridging positions – epistemically more defensible and pragmatically more useful than either extreme. Unfortunately, in defending compromise, the authors seem satisfied with "soft naturalism", a concept that is too amorphous to set useful conceptual structures for a much needed future nosology that successfully captures disorders with diverse etiologies and pathophysiologies, ranging from monogenic neurodevelopmental disorders associated with autism to highly polygenic disorders with significant developmental and environmental etiological factors, such as depression and post-traumatic stress disorder, for which there are no bright line boundaries with health. Indeed, it is treacherous to consider psychiatric disorders (e.g., all disorders included in the DSM-5) as a natural grouping that can be conceptually encompassed by any single definition or generalization. What is contained in the DSM-5 list of mental disorders is historically contingent, based in part on phenotypic similarities, in part on ignorance of etiology and neural mechanisms, and importantly on disciplinary claims and practice patterns that – as illustrated by movement of disorders among chapters in the ICD-11 – may be contested by other disciplines, such as neurology, sex endocrinology or pediatrics2, 3. Philosophy is a highly abstract discipline when focused on basic principles such as being, knowing, causality, and basic justifications for ethical principles. In contrast, as well illustrated by Stein et al, philosophy of psychiatry is concerned with applications of philosophical tools to the complex brew of theories, cognitive schemata (including disease classifications), hypotheses, empirical observations, policies and practices of a particular medical discipline. Philosophy of psychiatry gains traction on issues of importance to its diverse stakeholders through its interdisciplinary stance, with one foot in basic philosophical concepts and analytical methods, and the other firmly planted in the theoretical constructs, practices and science that undergird the discipline. In this context, the paper by Stein et al is especially timely, because the scientific basis of psychiatry is evolving rapidly and is already changing some corners of practice and policy – developments that will clearly benefit from philosophical analyses. During the last two decades, significant advances in technology and computation (including machine learning and data science) have not only opened new scientific possibilities, but have also motivated significant changes in the organization of biomedical science, such as the creation of international data-sharing consortia (e.g., the Psychiatric Genomics Consortium4) and large deeply phenotyped and genome sequenced populations cohorts (e.g., UK Biobank). Psychiatric research has exploited these developments to produce remarkable advances in genetics4; progress in relevant aspects of molecular neuroscience, connectomics and cognitive neuroscience; and powerful new approaches to phenotyping and epidemiology based on electronic medical records, large recontactable cohorts, and data sciences. Only fifteen years ago there were zero replicable genetic associations with psychiatric disorders. Today there are hundreds, and a few examples of integrated genetic and neurobiological data have already yielded candidate mechanisms for schizophrenia and other illnesses5. Such mechanistic hypotheses are in turn informing new rational discovery efforts for biomarkers and therapeutic targets. In the translational realm, connectomics and machine learning have produced circuit-based therapies (such as open- and closed-loop deep brain stimulation) that are already in clinical trials or in clinical use6. Growing evidence for brains as "prediction machines", emerging from cognitive and systems neuroscience, makes a strong case that brains construct subjective experience and behavior from Bayesian priors corrected by interoceptive and exteroceptive sensory information. If such understandings of brain function continue to gain evidentiary and conceptual support, as I believe they will, ideas about brain-mind relationships, sense of self, personal identity, and agency – so central to philosophy and psychiatry – will require substantial revision. In terms of mental illness, a Bayesian brain model can provide useful new ways for thinking about conditions as diverse as addiction, hallucinations and delusions7, and the mechanisms by which cognitive-behavioral psychotherapies exert beneficial effects. All of these scientific developments have profound philosophical implications not only for the classical problems of philosophy of psychiatry, but also for a range of issues ranging from the concept of self to volitional control of behavior. Modern genetics and neuroscience provide much information that should influence modern attempts to define psychopathology or classify psychiatric disorders. Human genetics demonstrates that DNA sequence variants associated with psychiatric disorders ("risk alleles") are continuously distributed in populations, without any break or inflection point that would naturally demarcate a categorical separation from health2. Epidemiological studies also demonstrate the continuous distribution in the population of symptoms and impairments associated with major depressive disorder and other common psychiatric conditions. In addition, individual risk alleles act pleiotropically, influencing multiple disorders. For example, schizophrenia and bipolar disorder share approximately 70% of their common variant risk alleles. For disorders that have been studied, unbiased single cell analyses of gene expression (from sequencing RNA) and epigenomics (from patterns of chromatin accessibility) performed in post-mortem brains find only quantitative differences between affected and unaffected individuals, and much overlap in gene expression between the disorders. These findings impugn the setting of categorical boundaries between DSM disorders, and strongly favor transdiagnostic dimensional definitions. Moreover, significant "within disorder" heterogeneity of underlying mechanisms confounds clinical research, contributing to the failure of attempts, through structural and functional imaging, to robustly and reliably distinguish cases from controls. Thus, current categorical disorder definitions not only impose artificial boundaries between named disorders, but also fail to yield homogeneity within putative categories. One additional discovery deserves emphasis here: at the genetic as well as the phenotypic level, diagnosed disorders do not exclusively represent deficits. For example, obsessive-compulsive disorder, polygenic autism, and anorexia nervosa share alleles with beneficial phenotypes, such as greater educational attainment and, in the case of autism, greater cognitive ability8, 9. Overall, a significant number of DNA variants associated with impairing symptoms may also be associated with strengths that have opposite valences with respect to life success and survival. Such scientific observations certainly complicate, if they do not defeat entirely, attempts of commentators to claim that psychiatric disorder definitions can be based on their speculative inferences about natural selection. Psychiatry is entering a time of significant change in its underlying science, that is just starting to influence the clinic6 and to gain the attention of policy makers. This is a time in which the analyses and reflections of philosophy of psychiatry should gain in importance, driven by engagement with new science. Stein et al have done an important service in providing a lucid platform from which psychiatry can better address its future.