A multilayer ontology of psychological stress ranging from lived experiences to neurobiological cascades
Year: 2026
Type: article
Abstract: Chronic stress is a phenomenon measured in pieces: an appraised predicament, a behavioral pattern, an autonomic shift, an HPA-axis cascade, an inflammatory-metabolic state, a molecular signature. Each is well characterized, but they are rarely reconciled in one computable model. We present the Corticoidogenic Stress Frames (CSF), a modular ontology network in which eight conceptual frames are each anchored to a primary corticoidogenic literature. CSF formalizes eight stress-related strata with their pathways and the cascade steps occurring at each layer. Appropriate competency questions validate the result on knowledge graphs extracted from stress-related datasets. CSF-shaped knowledge graphs are extracted automatically either from physiological data or linguistic reports. The extraction is performed by POLANYI++, a neurosymbolic AI pipeline in which a LLM recognizes patterns, while the ontology supplies the inductive bias, the vocabulary and the validation. Sensor features and first-person narrative bind to shared cascade steps, so quantified physiology and appraisal data are represented in a graph, while keeping their provenance. We present six studies. Study 0 confirms on a signal-free dataset that the pipeline returns a null rather than fabricating structure. Study 1 shows CSF constructs are sensitive to known conditions across three benchmarks. Study 2 finds a within-subject coupling between daily activity and momentary stress, positive within individuals, yet dissociable from the protective trait-level exercise buffer of Study 1. Study 3 evaluates extraction from forty-one unannotated diary posts by a reverse method, extracting the conceptual frames convergent with model-free affect signals. Study 4 shows that the co-activation distribution over multilayer frames is stable. In Study 5, five models from three families judge extraction items under blinding, aggregated by a Dawid–Skene latent-class model. By design each specialist rules only within its own stratum and abstains elsewhere, which keeps the judgments free of cross-expertise contamination. As an external human referent, a neuropsychologist rated a sample, endorsing the extracted frames in 11 of 13 cases. The framework offers a route to layer integration in stress physiology, with evidential status of each layer represented explicitly.
Source: Frontiers in Physiology
Authors: Chiara Lucifora(Conceptualization, Data curation, Formal analysis, Methodology, Writing – original draft, Writing – review & editing), Massimo Silvetti(Formal analysis, Validation), Aldo Gangemi(Conceptualization, Data curation, Formal analysis, Methodology, Software, Writing – original draft, Writing – review & editing)
Language: English
Subfield: Behavioral Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: gold
Estimated APC paid: $3,801
OpenAlex ID: W7236285287
