nucleation

THEORY-285: In an illustrative active-inference simulation of post-traumatic context inference, run once per phenotype, a prior that one is in a harmful context, made overly precise by a run of harmful trials under the "vulnerable" learning and forgetting weights, makes exploratory sampling look uninformative, so the agent attenuates the periphery, never receives the evidence that would correct the prior, and the belief sustains itself; lowering precision makes the attenuation policy less vigorous without correcting the prior, and the policy returns when precision is restored

4 min read

Properties7
StatusActive
Version2
Tagspsychopathology-and-treatment, neuroscience, complex-systems
SourceLIT-983 — Impaired context disambiguation as a key determinant of post-traumatic psychopathology: A neurocomputational account
Bears onQUESTION-045 — Is a mental disorder that outlasts its trigger a protective lock-in at the individual scale, a regime kept stable by behaviour that blocks its own disconfirmation while the capacity to leave it degrades, and does cutting the maintaining link, with the regime's content held fixed, produce lasting change?, QUESTION-303 — After cued traumatic recall, do people with persistent PTSD sample disambiguating peripheral context less than matched trauma-exposed controls, and does restoring that sampling during the same exposure produce lasting change that the exposure alone does not?
SupportsCLAIM-339 — A judgement–observation loop cut off from action loses one route for evidence it did not select, and the free-energy principle's reason for minimizing surprise, that a persisting system must keep its states in bounds, needs action on the world, though a loop can confirm itself with action available too, CLAIM-342 — What keeps a judgement–observation loop from confirming itself is evidence whose arrival or scoring does not pass through the interpretation it tests; action is one source of such evidence, not a necessary one, and a loop that selects its own evidence can lock in with action available, CLAIM-446 — Some persistent anxiety may be governance-maintained rather than belief-maintained: anticipated threat grants defensive processes authority over action, attention and evidence acquisition; under the avoidance that authority selects, the hypotheses that the situation is dangerous and that it is safe predict the same observation, so repeated safety need not correct the threat estimate; relief reinforces the defense without being evidence of danger; and the regime can outlast a corrected belief, its pathology, where present, being defensive authority insensitive to corrective evidence at avoidable cost
Cited as “extends” byTHEORY-1179 — In an active-inference model of post-traumatic context inference, an overly precise prior that one is in a harmful context makes exploratory sampling look uninformative, so the agent attenuates the periphery, never receives the evidence that would correct the prior, and the belief sustains itself; lowering precision removes the symptoms without correcting the prior

Graph View

Source

Linson, Anderson, Markham, Friston & Moutoussis (2025), bioRxiv, LIT-983, version 2, read in NOTE-766: Overview, Results (Figs. 4–7), Discussion and Online Methods.

What was actually shown

The model. A partially observed Markov decision process with two hidden contexts, harmful and harmless. They share an ambiguous first observation and differ only in what the periphery shows. At each step the agent either samples the periphery (exploratory) or attenuates it and keeps to the foreground, which “generates the same observation irrespective of context”. Context priors are learned Dirichlet counts with forgetting. Policies are chosen by expected free energy, so the agent explores only when it expects exploring to resolve uncertainty.

The loop (Figs. 6, 7a). After a short run of harmful trials, a “vulnerable” agent’s harmful-context prior becomes precise enough that exploration is “falsely believed to be uninformative due to overly informative prior beliefs”. It selects sensory attenuation, learns nothing about the harmless context, and so keeps selecting attenuation: “no amount of observations will render prior beliefs obsolete”. A “resilient” agent, given the same trials but different learning and forgetting weights, keeps exploring and rebalances. By trials 50–60 the vulnerable agent attenuates “under any condition”.

The interventions (Fig. 7b–f). Sixty harmless-only trials restored exploration near the end of the block, and the priors rebalanced and stayed balanced. Lowering the precision of the A and B tensors under random trials made policy selection less vigorous, read as fewer symptoms. But the priors did not rebalance, and when precision was restored the attenuation policy returned even on harmless trials.

What could have come out otherwise: exploration could have been chosen anyway, if the information gain from the periphery outweighed a confident prior; the drug arm could have let enough evidence through to rebalance. Within the chosen parameters, neither happened.

What this does not say

  • It does not show that PTSD works this way. Nothing is fitted to patient data. The vulnerable and resilient phenotypes are defined by learning and forgetting weights that the main text does not report. The match to the EMDR–fluoxetine trial (van der Kolk et al. 2007) is asserted by citation.
  • It does not show the loop is inevitable. Whether it forms depends on parameters. The paper shows one run per phenotype, not where in parameter space the tipping point lies.
  • It does not isolate the maintaining link. The exposure arm is harmless-only and the drug arm is random, so the arms differ in environment as well as in what they act on. The comparison QUESTION-045 asks for, the same exposure with and without restored sampling, is not run.
  • It does not test the circuit. The acetylcholine–noradrenaline and serotonergic-brake account, and the flashback account, are narrative and do not drive the simulations.
  • It is about version 2. The current version 3 (2026-05-13) was not read.

Connections

  • QUESTION-045 asks whether a disorder that outlasts its trigger is a protective lock-in, kept up by behaviour that blocks its own disconfirmation while the capacity to leave it degrades. This account gives the structure in model form: attenuation is the blocking behaviour, and the collapse of two basins to one is the lost capacity to leave. It answers the “is it” half only for a model. The “what would show it” half is answered by the paper’s predictions (peripheral against focal attenuation after cued recall), not by its results.
  • THEORY-058 holds that disorders are kept up by self-reinforcing loops and outlast their triggers. This is a mechanism for one such loop, with the loop running through what the agent chooses to observe.
  • THEORY-053 says only change in the theorised process, ahead of the outcome, bears on a theory of change. The drug arm is a modelled instance of outcome without process, and it relapses.

Backlinks