Measurement-gap predecessor

### PDF page 16 The Measurement Gap in Coercive Psychiatry Page 14 of 16 Each of these is, in principle, obtainable. None has been provided. Until one is, the burden of justification — which in the case of coercion lies with those who would override consent — remains unmet. 12. Conclusion This paper has not claimed that antipsychotics never help, that psychiatry is a fraud, or that no one in crisis should ever be treated. It has claimed something narrower and, for that reason, harder to dismiss: that the specific practice of compelling people to take antipsychotic drugs rests on a benefit that the discipline’s own instruments can barely measure and its own metrics leave contested, while inflicting harms that are measured to the standards of physical medicine. The significance is not abstract. The people on the receiving end of this asymmetry are real, their numbers are large and rising, and the harms recorded against them — the irreversible movement disorders, the metabolic disease, the regulatory carcinogenicity signals — are not projections but counts. When a society removes a person’s right to refuse a medical treatment, it assumes the heaviest burden of proof known to medicine. The evidence assembled here, drawn almost entirely from psychiatry’s own record, indicates that in the case of compulsory antipsychotic treatment that burden has not been met. That conclusion ought to be uncomfortable, and it ought to prompt the one response the evidence supports: not the abolition of psychiatric care, and not a bar on genuinely time-limited emergency intervention, but the abolition of sustained coercive antipsychotic maintenance absent the evidence its compulsion requires, and the redirection of resources toward approaches a consenting person would choose. Declaration of Generative AI and AI-Assisted Technologies in the Manuscript Preparation Process During the preparation of this work, the author used generative AI tools (including large language model assistants) for literature-search assistance, source verification, editorial critique, and drafting support. After using these tools, the author reviewed, verified, and edited the content as needed, checked all cited figures against their primary sources, and takes full responsibility for the content of the article.

precondition that distinguishes measuring temperature from rating a film. The philosopher of measurement Joel Michell has argued, across a body of work in mainstream measurement journals since 2000, that psychometrics rests on the unexamined hypothesis that psychological attributes are quantitative — and that this hypothesis has never been critically tested, the failure to test it being obscured by an anomalous definition of measurement. He defines a “pathology of science” precisely: a hypothesis is accepted without serious attempt to test it, and that first-level failure is then ignored. On this account psychometrics is “neither a quantitative science … nor a rational science” (Michell, 2000, 2008). This paper does not need Michell’s strongest conclusion to be correct, and does not adopt it as a premise. His is a minority position within a live debate, and the argument that follows is deliberately a continuous latent trait rather than demonstrating it, and their conjoint formulations are fitted to data rather than tested by experimental manipulation of the attribute — the operation that conjoint measurement theory actually demands. The quantitative hypothesis is thus accommodated, not established. The reader who rejects this reply may discard Section 2 entirely; Sections 3 through 6 stand without it. 3. Link One: The Diagnosis That Defines Trial Entry Is Unreliable Every antipsychotic efficacy trial begins by classifying participants into a diagnostic category. If two

disorder and any improvement can be validly measured (M); that the drug confers a clinically meaningful benefit (B); that adding legal force improves outcomes over voluntary, supported care (F); that the serious risk said to justify compulsion is individually predictable (R); that the harms are proportionate to the benefit (H); and that no less restrictive alternative would be adequate (L). The preceding sections contest M, B, and H. This section and the next address F and R, which are the conditions the standard justification most conspicuously fails to establish — and the conditions on which compulsion most directly rests. against voluntary care, the burden that compulsion requires is unmet in both — by negative evidence in one case and by absent evidence in the other (Barnett et al., 2018; Burns et al., 2013; Kisely et al., 2017). 7.2 Condition R: individual risk is not reliably predictable Compulsion is frequently justified by the prediction that a particular person will, without forced treatment, come to serious harm or cause it. That prediction is carried by structured risk-assessment instruments, and their performance does not bear the weight placed on it. The most comprehensive synthesis — a 2012 BMJ systematic review and meta-analysis of seventy-three samples and nearly twenty-five thousand people — found that violence-risk tools discriminate only moderately (areas under the curve in the range of roughly 0.66 to 0.74) and, crucially, perform far better at identifying who is

sufficient on their own to justify decisions about detention or release. The asymmetry matters for coercion specifically: a tool that can clear the low-risk but cannot reliably confirm the high-risk cannot supply the individualised predictive certainty that overriding a person’s autonomy is supposed to rest upon. Compulsion premised on “this person is dangerous without treatment” is premised on a prediction the field’s own tools cannot reliably make at the individual level (Fazel et al., 2012). Conditions F and R are not peripheral. They are the two predicates that distinguish coercion from ordinary care, and they are precisely the two for which the supporting evidence is weakest. The stronger form of this paper’s thesis follows: sustained psychiatric coercion rests on two unproven individual-level predictions — that this person will suffer or cause serious harm without forced treatment, and that forced treatment will deliver a clinically meaningful benefit exceeding its harms and exceeding what less restrictive care would achieve. The existing literature establishes neither predicate to the standard that the removal of consent demands. 8. The Other Side of the Ledger: Harms on Biological, Regulatory, and Registry

outcomes to supported, consenting care (Barnett et al., 2018; Burns et al., 2013; Kisely et al., 2017). • A validated risk-assessment instrument able to predict serious violence or self-harm at the individual level with accuracy sufficient to justify detention — substantially exceeding the moderate discrimination (AUC ~0.66–0.74) of current tools, which perform far better at clearing low risk than confirming high risk (Fazel et al., 2012). Page 14 of 16