Temperature as Precision: A First-Person Account of Hypomanic Cognition through the Lens of Active Inference
Gayeong Song
Zenodo (CERN European Organization for Nuclear Research) March 31, 2026 preprint DOI: 10.5281/zenodo.19341743 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Commentary |
|---|---|
| Sample size | 1 |
| Population | The author, a computer science student with Bipolar II Disorder, reflecting on personal episodes |
| Key points | Proposes a three-axis model (Temperature, Valence, Volatility) to describe precision failures in Bipolar II Disorder, arguing that five retrospectively analyzed hypomanic thought trajectories converged on a single grandiosity attractor. Contends that precision reduction in hypomania is non-uniform, with question generation intact but answer validation collapsing, and that self-censorship and differential pharmacological responses can be interpreted as precision-related phenomena. |
Abstract
This commentary reinterprets the precision engineering framework proposed by Friston (2023) through the lens of first-person experience in Bipolar II Disorder (BP2). As a BP2 patient and computer science student, the author models cognitive fluctuations across hypomanic, mixed, depressive, and euthymic episodes using a Temperature (τ) framework, where τ ≈ 1/precision. The paper proposes a three-axis model ― τ (Temperature) × V (Valence) × σ (Volatility) ― to capture the phenomenology of precision failures. Five retrospectively analyzed hypomanic thought trajectories demonstrate convergence to a single grandiosity attractor despite diverse initial seeds. Key observations include: the non-uniform nature of precision reduction in hypomania (question generation remains intact while answer validation collapses); the self-censorship filter as a manifestation of precision-weighted prediction error suppression; and differential pharmacological responses interpretable as τ modulation versus output suppression. These observations suggest that precision engineering is not only a theoretical construct but a subjectively experienced phenomenon, and that computational psychiatry may benefit from systematic integration of first-person autoethnographic data.