Isolating the exposomic impact of acute war and displacement within chronic crisis using multimodal, AI-enabled phenotyping

Wars often unfold within contexts of prolonged instability, yet little is known about how acute conflicts interact with cumulative exposomic adversity to shape mental health processes. Here we studied clinical and subclinical symptoms in a Lebanese sample across three groups—one assessed before the 2024 war and two assessed after, with or without displacement—within a shared background of chronic crises. Using a multimodal, AI-enabled approach combining clinical assessments, narrative coding, and large-scale linguistic and emotional profiling of life stories, we show that mental health burden does not scale linearly with acute war or displacement. Despite higher cumulative exposure, war-time groups did not exhibit higher PTSD or depression than pre-war participants, and resilience was lowest among non-displaced individuals. In contrast, narrative organization showed non-linear reconfiguration: displacement was associated with enhanced temporal structuring but reduced global causal coherence, whereas acute war exposure without displacement amplified autobiographical reasoning and explicit self-change. AI-based analyses further revealed emotional blunting rather than emotional overload, alongside increased linguistic complexity decoupled from symptom severity. By enabling scalable analysis of life narratives, this approach isolates acute vs chronic effects and reveals sub-clinical adaptations not captured by symptom measures alone, that can inform healthcare and policy in conflict-affected regions.