The Algorithmic Architecture of Amphetamine-Induced Psychosis
Reframing stimulant psychosis from a static "chemical imbalance" into a dynamic failure of biological inference. When amphetamines hijack precision weighting, random sensory noise becomes hyper-salient and rigid top-down priors overwrite reality.
The molecular & circuit cascade: from VMAT2 reversal to sensitization
Amphetamines penetrate monoaminergic neurons, reversing VMAT2 to empty synaptic vesicles into the cytosol and forcing reverse transport of dopamine via DAT. The surge triggers cortical glutamate overflow, damages GABAergic interneurons, and erodes prefrontal inhibitory control over subcortical structures.
VMAT2 & DAT reversal
Amphetamine forces cytosolic dopamine accumulation and non-exocytotic efflux through dopamine transporters into the synaptic cleft, independent of action potential firing.
Glutamatergic excitotoxicity & GABA loss
Unconstrained dopamine triggers cortical glutamate overflow, selectively damaging NMDA-modulated GABAergic interneurons in prefrontal cortex and destroying excitation/inhibition balance.
Endogenous sensitization
Repeated exposure permanently lowers the threshold for subcortical dopamine release. PET studies show mild amphetamine regimens induce surges resembling drug-naive first-episode schizophrenia.
Subcortical dopamine release profile
PET radioligand binding delta across sensitization states
Figure 1.1: progressive amplification of dopamine efflux with repeated exposure.
Hierarchical predictive coding engine: interactive Bayesian dynamics
Perception is a precision-weighted trade-off between top-down priors and bottom-up sensory data. Dopamine biophysically encodes precision. Amphetamines distort that gain, inflating precision signals and corrupting belief updating.
Interactive Bayes simulator
Adjust precision parameters to see how hyper-precision shifts the posterior belief curve in real time.
Bayesian belief updating distribution
Posterior perception is a precision-weighted compromise between prior expectation and sensory likelihood.
Dopamine flood assigns high precision to random noise, causing aberrant salience. The brain constructs paranoid explanations for hyper-salient noise.
Dopamine flood inflates the precision of top-down priors. Under ambiguity the internal model overwrites sensory input, generating hallucinations.
Algorithmic pathologies: quantitative behavioral clamps
Computational psychiatry uses task paradigms, behavioral clamps, to isolate specific failure modes in decision making, probabilistic reasoning, and sensory processing under stimulant exposure.
Jumping to conclusions (beads task)
Draws to decision across trait vulnerability and acute dopamine state
Figure 3.1: low baseline draws-to-decision plus acute dopaminergic load triggers rapid delusional crystallization (Ermakova et al., 2014).
Perceptual bias under uncertainty
Assimilation to the mean in temporal duration tasks (Cassidy & Horga, 2018)
Figure 3.2: high striatal dopamine induces overconfidence in priors when input is noisy.
Social cognition (10-trial trust game)
Methamphetamine users show marked theory-of-mind deficits, heightened irritation, and reduced guilt. Their algorithms fail to update predictions of partner intent, impairing long-term mutual returns.
Multi-step planning & aversive pruning
Under interoceptive stress, stimulant users prune decision branches containing immediate discomfort, failing to reach optimal long-term goals.
Clinical epidemiology, dose thresholds & trajectory
Risk of algorithmic destabilization is tightly linked to pharmacological mechanism and dose. Differentiating medicinal amphetamines, methylphenidate, and chronic misuse allows accurate clinical risk modeling.
Psychosis risk by dosing threshold
Amphetamine (30 mg/day threshold) vs. methylphenidate (Moran et al.)
Figure 4.1: exceeding 30 mg/day sharply increases psychosis risk; methylphenidate, a reuptake inhibitor only, shows no significant increase.
Clinical outcome & transition trajectory
Longitudinal breakdown of stimulant psychosis cases (Murrie et al., 2019)
Figure 4.2: most cases resolve within 1-2 weeks, 22% transition to schizophrenia, 8% take a persistent delayed course.
Differential diagnosis: amphetamine psychosis vs. primary schizophrenia
| Feature | Amphetamine-induced psychosis | Primary schizophrenia |
|---|---|---|
| Onset profile | Rapid, linked to acute binges or doses >30 mg/day | Gradual, preceded by an extended prodromal phase |
| Hallucinations | Vivid visual and tactile (formication) prominent | Predominantly auditory; visual/tactile less common |
| Formal thought disorder | Less prominent; cognitive organization preserved | Severe; loosening of associations and derailment |
| Autonomic signs | High (tachycardia, hypertension, mydriasis) | Typically absent in non-agitated states |
Systems dynamics & next-generation therapeutics
The sensitivity threshold model formalizes how cumulative neurochemical load pushes vulnerable systems past processing capacity into pathological attractor states. Treatment requires moving beyond crude D2 blockade toward precision circuit recalibration.
Sensitivity threshold calculator
Risk index = (sensitivity × load) / capacity
Classic D2 antagonists
Haloperidol blunts dopaminergic prediction-error signals. It treats delusional rigidity but indiscriminately dampens adaptive reward learning and worsens negative symptoms.
Muscarinic agonists (Cobenfy)
Xanomeline-trospium targets muscarinic receptors to restore cortical signal-to-noise indirectly, recalibrating subcortical dopamine without direct D2 blockade.
Metacognitive training (MCT)
Cognitive therapy designed to expose and correct algorithmic biases such as jumping to conclusions, retraining belief-updating alongside pharmacotherapy.
In-silico flight simulators
Computational multi-level models and large language models act as test beds to simulate individual reasoning failures and optimize personalized targets.