Using large language models to evaluate the interplay between posttraumatic stress disorder symptom severity and...

Intolerance of uncertainty (IU) is central to cognitive models of posttraumatic stress disorder (PTSD), reflecting a maladaptive process that maintains hypervigilant threat appraisals and symptom persistence. However, traditional self-report measures may overlook IU's implicit and context-dependent expression. Here, we introduce a novel language-based method to detect subtle linguistic markers of IU from naturalistic daily diaries and examined the role of IU in PTSD symptoms dynamics. In a 90-day ecological momentary assessment study, 432 trauma-exposed individuals with elevated PTSD symptoms provided 29,623 diary entries and 32,941 daily PTSD symptom ratings. A large language model (LLM)-based IU assessment quantified the similarity between daily transcribed speech and language from an established IU scale; IU model scores correlated with human ratings (r = .60). Dynamic structural equation models showed that higher daily IU predicted greater next-day PTSD symptom severity (β = .041, 95% credible interval CI [.024, .056]), even after controlling for prior-day symptoms. Higher daily PTSD symptom severity also predicted greater next-day IU (β = .035, 95% CI [.018, .054]). The strength of these bidirectional effects varied across individuals (IU→PTSD variance = .014, 95% CI [.007, .022]; PTSD→IU variance = .037, 95% CI [.030, .047]), reflecting substantial heterogeneity in daily IU-PTSD dynamics. Complementary multilevel analyses demonstrated that language-derived IU provides incremental validity, explaining variance in PTSD symptom severity beyond sentiment and positive and negative affect. These findings extend cognitive models of PTSD, demonstrating that IU influences daily symptom trajectories, and provide an ecologically valid, scalable framework for tracking implicit cognitive processes via natural language. (PsycInfo Database Record (c) 2026 APA, all rights reserved).