nucleation

LIT-983: Impaired context disambiguation as a key determinant of post-traumatic psychopathology: A neurocomputational account

2 min read

Properties5
StatusActive — read 2026-10-14 (NOTE-766) in full, in version 2 (2025-08-04); version 3 (2026-05-13) could not be reached, as bioRxiv returned HTTP 429 to every request. Worth reading as a worked mechanism of a disorder that keeps itself in place: under active inference an overly precise prior for a harmful context makes exploring the periphery look uninformative, so the agent attenuates it, never gets the corrective evidence, and the belief sustains itself; exposure that restores sampling rebalances it, and a precision-lowering drug only suppresses symptoms. Read it knowing that every result is an illustrative simulation with unreported parameters, and that its two intervention arms differ in environment as well as mechanism. THEORY-285 holds what it shows.
Version2
Tagspsychopathology-and-treatment, neuroscience
Cited as “paper” byNOTE-766 — Linson et al. — Impaired context disambiguation in post-traumatic psychopathology
Cited as “source” 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, 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

Graph View

Adam Linson, Heidi Anderson, Sarah Markham, Karl Friston and Michael Moutoussis (2025), bioRxiv preprint.

Key takeaways

From the reading, NOTE-766 (version 2).

  • The maintaining link is a choice of what to sample. A harmful and a harmless context share an ambiguous cue and differ only in the periphery. With a precise harmful-context prior, exploration is “falsely believed to be uninformative”. The agent attenuates the periphery, and “no amount of observations will render prior beliefs obsolete”.
  • Vulnerability is a learning rate. Resilient and vulnerable agents differ in “the specific weights of learning and forgetting”. After the same trauma, one rebalances and the other collapses into a single basin. Parameter values are not given.
  • Exposure against a drug. Sixty harmless-only trials restore exploration and the priors rebalance durably. Lowering likelihood and transition precision, standing in for SSRIs or alcohol, suppresses symptoms without rebalancing, and they return when precision is restored. The arms also differ in environment (controlled against random).
  • The rest is narrative. A pontine acetylcholine–noradrenaline loop under a serotonergic “brake”, and flashbacks as top-down context in visual working memory, are proposed from the literature and not modelled.
  • Stated tests. Cued recall should attenuate peripheral more than focal sensing. Context should modulate startle less in PTSD. Active inference and dual representation theory predict opposite weights on trauma-encoded context.

Standing in the record

Filed on 2026-10-12 for QUESTION-045, and read on 2026-10-14 in NOTE-766, in version 2. It is what the filing said: an individual-scale model of a regime kept stable by behaviour that blocks its own disconfirmation, with a lost capacity to leave. It does model the maintaining link and not only the content. The drug arm removes the symptoms but leaves the link, and relapses. But it is a model. The intervention arms are not matched on exposure, so it does not run the comparison QUESTION-045 asks for. THEORY-285 holds the mechanism at simulation strength. It gives THEORY-058’s loop a mechanism, and its drug arm is an instance of THEORY-053’s outcome-without-process. Version 3 should be read when bioRxiv can be reached, to check that the model and results are unchanged.

Backlinks