Michael Darius Eastwood · Independent AI alignment researcher
Published
The Capability-Adjustment Fallacy is the institutional inference that visible high performance in one domain constitutes evidence that reasonable adjustments are unnecessary in another. The same neurodivergent trait can be simultaneously an advantage and an impairment, depending on context. Institutional inference that the advantage cancels the impairment is unsafe. The fallacy is the operational counterpart of the Polymathic Neurodivergent Profile (PNP) and is the mechanism by which adaptive-functioning deficits are dismissed as 'organisation', 'attitude', or 'not trying hard enough'.
The fallacy has a specific structural signature. An assessor observes evidence of high-tier performance in a chosen high-interest domain (research output, complex written work, cross-disciplinary synthesis) and reasons that the person could apply the same capacity to the low-tier domain (routine administration, timekeeping, task-switching) if they chose to. The reasoning ignores the well-documented executive-function asymmetry of the underlying condition: capacity does not port across the low-stimulation boundary, and effort applied to procedural work does not draw on the same substrate that produced the high-tier output. Adjustments that would be granted to a person without the visible high-tier evidence are withheld from the person who has it, because their strengths are read as counter-evidence to their needs.
The Capability-Adjustment Fallacy is named and defined in Paper C (Polymathy and Neurodivergent Cognition), the PNP working paper. It is presented alongside operational criteria for distinguishing PNP from ordinary high openness, ordinary giftedness, and generic executive dysfunction.
The fallacy is consistent with existing clinical guidance. NICE CG142 directs professionals to recognise the discrepancy between intellectual functioning (IQ) and adaptive functioning; the fallacy is a specific institutional failure that guidance is intended to prevent. Modern strengths-based assessment treats the same neurodivergent trait as advantage-or-impairment depending on context. Convergence 18 in the register (Hernandez-Espinosa, Abrahao, Witkowski, Zenil, PNAS Nexus 5(4) pgag076) identifies neurodivergent cognition as a protective factor for AI alignment: different minds catch what uniform minds miss.
Descriptive framework, not a diagnostic category. The paper is explicit: polymath is not a diagnostic label; PNP does not confer superiority; capability asymmetry does not remove the need for clinical assessment; PNP alone is not sufficient for legal or institutional decision-making. The framework is falsifiable: it must add incremental predictive value beyond existing autism and ADHD formulations, or it collapses into them.
From the book Infinite Architects: Intelligence, Recursion, and the Creation of Everything by Michael Darius Eastwood.