Neuropsychiatric Conditions with KT (AIT) Mapping

A reference table that maps 30+ neuropsychiatric conditions onto a three-part computational model of the mind, showing which "module" breaks down in each disorder.

The core idea is simple: treat the brain as an algorithmic agent with three components — a valence/objective function (what the agent wants, its reward signal), a world model (its generative, predictive beliefs about reality), and a planning system (executive control, policy selection). This framing comes from Ruffini & Castaldo 2024. The table then asks, for each disorder: which of these three is primarily disrupted?

The mapping produces some clean intuitions. Schizophrenia is mainly a world-modeling failure — hallucinations and delusions arise from mis-weighted generative predictions, while negative symptoms reflect planning deficits. MDD, by contrast, is centered on the objective function: anhedonia is a broken reward signal, pessimistic priors are a secondary downstream effect. OCD looks like a planning loop that can't terminate, driven by overestimated harm in the world model. ADHD is almost purely a planning deficit, with world modeling largely intact. Autoimmune encephalitis (anti-NMDAR) is the only condition rated primary across all three — a pan-domain collapse, which matches its clinical severity.

The table covers a wide range: classic psychiatric disorders (schizophrenia, bipolar, MDD, PTSD, OCD, eating disorders, addiction), neurodevelopmental conditions (ASD, ADHD, Down syndrome), neurological diseases (Alzheimer's, Parkinson's, Huntington's, TBI, epilepsy), and rarer entities like PANDAS and REM sleep behavior disorder. Each row includes a one-to-two line mechanistic note grounded in cited clinical literature, so the computational interpretation is tethered to empirical sources rather than floating free.

This is a working reference document, not a research paper — it is explicitly a "computational interpretation layered on clinical evidence," and the authors acknowledge that individual phenotypes vary. Its value is organizational: it gives clinicians and researchers a shared vocabulary for asking where in the agent's architecture a given condition intervenes, which could sharpen hypotheses about treatment targets or cross-disorder commonalities. The scope note flags several conditions not yet included, suggesting this is a living document intended to grow.