Psychological Non-Standardness in Top Literary Authors: A Computational Biographical Analysis of 567 Writers Across Three Centuries

We present a large-scale computational biographical study of 567 top-rated literary authors (Goodreads, N = 600) spanning three centuries (XIX–XXI), annotated across 38 predefined psychological tags plus an open-ended custom tag set. Using a standardness score (0–9 ordinal scale) annotated by a large language model from Wikipedia biographical text, we find that non-standardness is pervasive: median = 5.0 (IQR 3, 6), with 50.3% of authors scoring ≥5 and 23.5% scoring ≥7. The five most prevalent traits are childhood trauma (63.4%), chronic illness (45.8%), non-traditional relationships (44.5%), self-destructive patterns (36.1%), and depression (34.7%). Non-standardness is significantly lower in the XXI cohort vs XIX and XX (r = 0.38–0.42, p < 0.001), attributable to documentation bias. Tag-based logistic regression predicts high score with CV ROC-AUC = 0.941 ± 0.015; biography embeddings (nomic-embed-text-v1.5, 8192 tokens) achieve CV AUC = 0.787, confirming psychological signal in raw text. TF-IDF analysis reveals era confounding in text-based prediction. Mediation analysis shows a suppression effect: childhood trauma’s impact operates primarily through depression (109.8%) and self-destructive patterns (127.2%). The dominant custom tag is mortality-driven urgency (n = 115).