nucleation

THEORY-897: In Brown, Price and Dombrovski's reading of seven computational learning studies they selected without a systematic search, with mixed directions, anxiety goes with altered adjustment of learning rate to uncertainty, mostly for aversive outcomes, while overall learning rate, choice consistency and outcome valuation show little relation to it

3 min read

Properties7
StatusActive
Version2
Tagspsychopathology-and-treatment, learning-and-conditioning
SourceLIT-1479 — Anxiety as a disorder of uncertainty: implications for understanding maladaptive anxiety, anxious avoidance, and exposure therapy
Bears onQUESTION-329 — In people with fear-based anxiety disorders against controls, does a systematic search with meta-analysis, or a large preregistered study, find that what differs is the adjustment of learning rate to the kind of uncertainty, noise against volatility, mainly for aversive outcomes, while overall learning rate, choice consistency and outcome valuation do not?
CorroboratesCLAIM-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-1182 — Across seven computational learning studies selected without a systematic search, anxiety goes with altered adjustment of learning rate to uncertainty, mostly for aversive outcomes, while overall learning rate, choice consistency and outcome valuation show little relation to it

Graph View

Source

Brown, Price & Dombrovski, LIT-1479, read in full in NOTE-1120 from the NIH author manuscript (PMC10475148): the sections “Reinforcement learning studies of anxiety” and “Anxiety as maladaptive aversive uncertainty learning”, and Table 2. The version of record could not be fetched.

What was actually shown

The authors tabulate seven studies that model learning under uncertainty in samples with some clinical-level fear-based anxiety (Browning et al. 2015; Brown et al. 2018; Homan et al. 2019; Gagne et al. 2020; Norbury et al. 2021; Abend et al. 2022; Zika et al., preprint). Across them:

  • Differences lie in parameters for how learning adjusts to uncertainty (associability, volatility adjustment, latent-state formation). “Other parameters measuring overall learning rate, choice consistency, and outcome valuation show little relationship with anxiety.”
  • Studies with reward and loss conditions found differences in loss learning only (two) or in both (one).
  • Under high volatility, anxious people learn slightly slower overall but accelerate after very surprising outcomes; under high noise they learn faster, especially after surprising losses, which the authors say is less consistent.

What could have come out otherwise: anxious participants could have shown higher learning rates for punishment across the board, or higher valuation of losses, which is what an excess-threat account expects. In these studies, by the authors’ reading, they mostly did not.

What this does not say

  • That the literature was surveyed. The studies were chosen from the authors’ knowledge and searches, and the authors say they “do not represent a comprehensive, structured literature review following PRISMA guidelines”.
  • That anxiety is a failure to tell noise from change. That is the authors’ reading of the pattern. They list two rival readings: that anxious people infer more information from negative outcomes (Pulcu & Browning 2019), or underestimate noise (Piray & Daw 2021).
  • That the finding is specific to anxiety. One study found it more related to general internalizing symptoms than to GAD or depression; a computational meta-analysis they cite found higher punishment learning rates in both anxiety and depression.
  • Anything about avoidance. These are learning tasks without avoidance. The link to avoidance is the authors’ proposal, filed separately as THEORY-910.

Connections

It bears on CLAIM-446 as background: the anxiety it describes is a disorder of how evidence is weighed, not of how dangerous things are judged to be, which fits the claim’s statement that the defining feature “need not be a high estimate of danger”. It does not bear on authority.

Backlinks