nucleation

LIT-1496: Anxiety, Avoidance, and Sequential Evaluation

2 min read

Properties5
StatusActive — read in full 2026-10-24 (NOTE-1136) from the CC BY version of record; worth reading for the derivation that avoidance far from a threat needs no overestimate of the threat, only an expectation that one will fail to avoid it later, which spreads threat value backward across a decision process. Simulations only, matched qualitatively to published findings (earlier cash-out, behavioural inhibition, aversive pruning); the predicted loss of the free-choice premium is untested, and the persistence of avoidance after the misbelief is corrected is argued, not simulated.
Version3
Tagspsychopathology-and-treatment, learning-and-conditioning
Cited as “paper” byNOTE-1136 — Zorowitz, Momennejad & Daw — Anxiety, avoidance, and sequential evaluation
Cited as “source” byTHEORY-1190 — In a Markov decision process, an agent that evaluates actions as if it will fail to avoid danger at later steps, while in fact choosing optimally at each step, spreads threat value to states far from the danger and avoids early and widely, which reproduces qualitatively exaggerated threat appraisal, fear generalization and several laboratory findings associated with anxiety, THEORY-915 — In Zorowitz, Momennejad and Daw's simulations of a Markov decision process, an agent that evaluates actions as if it will fail to avoid danger at later steps, while in fact choosing optimally at each step, spreads threat value to states far from the danger and avoids early and widely, which reproduces qualitatively exaggerated threat appraisal, fear generalization and several laboratory findings associated with anxiety

Graph View

Zorowitz, S., Momennejad, I. & Daw, N. D. (2020). Anxiety, avoidance, and sequential evaluation. Computational Psychiatry 4: 1–17. doi:10.1162/cpsy_a_00026. Preprint: bioRxiv, doi:10.1101/724492 (2019).

Key takeaways

From the reading, NOTE-1136. Page numbers are the journal’s.

  • The core observation. “If danger can be avoided later, it poses less threat now” (p. 1): under optimal evaluation avoidable threats do not propagate to earlier states.
  • The bias. A β-pessimistic Bellman equation: the agent chooses optimally but values each action as if it will act badly later. “Importantly, this belief is false under the model” (p. 4). The danger itself is represented correctly.
  • What it produces. Threat value spreads across the state space; avoidance starts early and wide; earlier cash-out in the balloon task, slower approach under threat and aversive pruning follow, matched qualitatively to cited findings (pp. 3–7).
  • Persistence. “Frustratingly, these biased estimates of value may still remain even after a misbelief in the efficacy of future action is corrected for in a course of therapy” (p. 8), argued by analogy to model-free relearning, not simulated.
  • A prediction. Anxious people should show a reduced free-choice premium (p. 9), untested.
  • Scope. Prospective anxiety (generalized, social); not OCD compulsions or PTSD memory (pp. 11–12).

Why this is here

The exchange cited it at turn U14 for a computational account of anxious avoidance in which a pessimistic belief about one’s own future control spreads avoidance backward in time. The abstract matches.

Corrected 2026-10-24, from the reading. The exchange’s description is accurate; “backward in time” is, in the model, backward across antecedent states. The reading adds that the paper has no data of its own. For CLAIM-446 it is a rival account with the threat represented correctly and a belief about one’s own future control doing the work, so a test that holds only the threat estimate fixed cannot separate the two; and its explanation for avoidance outlasting a corrected belief is cached value, the habit-with-memory alternative the claim’s defeater names. Filed from it: THEORY-915.

Backlinks