Is precision psychiatry self-evident?

We believe their reply is important as it forms part of what Braslow and colleagues refer to as ‘psychiatry's taken-for-granted, everyday beliefs’: that the promise of precise biology to remedy mental suffering is enough to make it self-evident (Braslow, Brekke, & Levenson, Reference Braslow, Brekke and Levenson2020). The implicit premise of precision psychiatry is that phenomena of the mind are physically represented and that these representations are relevant to our understanding of mental suffering. This belief is so strong that it does not require explicit reflection, let alone further examination. To belong to the traditional academic psychiatric community is to reiterate the self-evident nature of the belief. To seriously entertain the hypothesis that, for example, schizophrenia may not be a self-evident disorder of the brain is dismissed as ‘antipsychiatry’ (Sommer, Kahn, Denys, Schoevers, & Aleman, Reference Sommer, Kahn, Denys, Schoevers and Aleman2015).

However, not everybody agrees with such a stance. In fact, psychiatry is faced with increasingly blunt assessments of its reductionist belief system, written by eminent – mostly non-psychiatric – scholars in prestigious mainstream journals (some examples in Table 1). Since the civil rights movements of reformist psychiatry in the 1970s (later framed as ‘antipsychiatry’) and the ‘recovery movement’ in the 1980s, the patient's voice, and the focus on the existential domain of personal recovery, has not gained much traction in academic psychiatry, some even considering it a ‘hoax’ (Schizophrenia Research Forum, 28 November 2017). As a result, recovery-focussed work in the mental health care sector is hampered by limited ‘institutional readiness’ (Leamy et al., Reference Leamy, Clarke, Le Boutillier, Bird, Janosik, Sabas and Slade2014) and by psychiatry's inability and/or unwillingness to escape the ‘epistemic prison’ of the ‘right medication for the right DSM-diagnosis’ (Gardner & Kleinman, Reference Gardner and Kleinman2019; Hyman, Reference Hyman2010). In reaction, novel and increasingly popular movements like Mad in America (MIA; www.madinamerica.com) produce a stream of articles focussed on debunking the same belief system that Salagre and Vieta passionately advocate. Although it is easy to dismiss MIA as unreasonable ‘antipsychiatry’ (Sommer et al., Reference Sommer, Kahn, Denys, Schoevers and Aleman2015), it may also be viewed as a campaign to address epistemic injustice (Crichton, Carel, & Kidd, Reference Crichton, Carel and Kidd2017) brought about by the ‘myopia’ (Braslow et al., Reference Braslow, Brekke and Levenson2020) of an academic psychiatry that has unlearned to hear an individual's personal meaning in the experience of mental distress.

Table 1. Recent publications about how ‘precision psychiatry’ is viewed outside psychiatry

AI solutionism

Apart from raising the model of pathological-brain/body-causes-abnormal-mind to the level of self-evidence, ignoring the ‘hard problem of consciousness', Salagre and Vieta assume another self-evident property of mental suffering: that it is determinable and predictable. Thus, they assume that there naturally will be such a thing as the ‘right treatment’ at the ‘right dose’ at the ‘right time’ for an individual with mental distress. In their view, analytical innovations such as machine learning will unravel determinability and predictability for use in clinical practice. However, this stance overlooks the underlying question to what degree mental suffering can be considered determinable and predictable in the first place – and therefore runs the risk of becoming a form of messaging called ‘AI solutionism’ (AI = artificial intelligence). This is the supposition that, as long as there is enough data, any human outcome can be computed based on machine learning algorithms (Chen & Asch, Reference Chen and Asch2017). The question, however, is whether this can be considered a reasonable hypothesis in the case of mental outcomes.

AI solutionism holds that mental outcomes are determined – and therefore predictable. However, mental outcomes are unpredictable because they are inextricably linked to stochastic events in a complex system where chaos theory rules. Chaos theory describes the phenomenon that with the wisdom of hindsight all events may well be determined – but prospectively remain unpredictable. Although there are undoubtedly factors which have a statistical association with mental outcomes – for example, the loss of a loved one and the subsequent feeling of sorrow – the moment of losing a loved one at some point in time is unpredictable.

It is attractive to assume that, say, getting better on an antidepressant is a kind of determined process so that machine learning based on ‘everything’ can predict treatment response in a particular patient. However, the scientific basis for this is lacking, both conceptually and meta-analytically (e.g. Kennis et al., Reference Kennis, Gerritsen, van Dalen, Williams, Cuijpers and Bockting2020). The reason for this was described by Tikhodeyev and Shcherbakova in the context of the mutagenic effect of ultraviolet radiation. Although the amount of mutagenic damage in microorganisms can be reliably predicted based on the amount of radiation, temperature, duration and culture medium, it cannot be predicted in which microorganism and where, in this microorganism, in the genome mutations will occur (Tikhodeyev & Shcherbakova, Reference Tikhodeyev and Shcherbakova2019). The explanation, according to chaos theory, is that even in the case of a deterministic (non-random) process, simple nonlinear systems cannot be predicted in the future. Machine learning cannot solve this (Chen & Asch, Reference Chen and Asch2017).

The same applies to mental suffering: although there are weak therapeutic influences of factors at the group level, it remains unpredictable whether these influences lead to change in the stochastic ecosystem of a specific individual. Unpredictability in the ecosystem is partly due to the so-called butterfly effect: the sensitivity of the future to the most minute random change in the baseline condition. The importance of the butterfly effect is all the greater when one understands that ‘getting better’ on an antidepressant is largely dependent on a complex placebo-effect that has to do with expectation, relationship, being observed and time (Kirsch, Reference Kirsch2014). Therefore, we fail to see why determinability and predictability should be considered self-evident postulates of precision psychiatry.