Treating The Prodrome

An illuminating mechanistic case for treating the schizophrenia prodrome. The 'Duration of Untreated Psychosis' model holds psychosis is 'toxic' (brain-damaging), so early intervention prevents full schizophrenia — but the evidence is mixed (confounded observational studies; failed early-intervention RCTs), and Scott was previously skeptical because it's so convenient for the pharma-aligned 'treat aggressively' camp. After learning the biology he's more sympathetic, via a predictive-coding model (Friston; Fletcher & Frith; Surfing Uncertainty): schizophrenia starts with aberrant prediction errors (a buzzing fly feels cosmically significant); early ones are dismissible, but as they accumulate the brain RATIONALLY shifts high-level beliefs toward psychotic worldviews (signs from God, conspiracies) in a feedback loop, eventually twisting its highest parameters into 'evidence is random, don't bother' -> negative symptoms (amotivation, catatonia). Antipsychotics stop the prediction-error spikes (positive symptoms) but can't untwist beliefs already locked in over years — which is exactly why DUP would matter. Vividly grounded in Peter Chadwick's first-person account of his cascade into delusion and bus-suicide attempt, then recovery on antipsychotics.

Quality 74 · Strong. Strong, high (74). A genuinely illuminating synthesis — applying the predictive-coding / Bayesian-brain framework to explain why schizophrenia progresses and why early treatment plausibly matters — well-grounded and vivid (the Chadwick account). Held just below Excellent by its technical niche and Scott's appropriate hedging on DUP itself.

Claude’s paradigm shift 58 · Moderate. Moderate, high (58). Draws on Friston/Frith's framework, but applying predictive coding specifically to the DUP/prodrome question is Scott's fresh synthesis.

Real-world impact 2 · Minor. Minor (2). An influential treatment within the psychiatry/computational-neuroscience niche, but one post with no direct policy/practice effect. (Predictive-coding model is an active research program, not settled, but it is neuroscience, not social psych -> no replication caveat.)