racing Personality Structure in Narratives: A Computational Bottom–Up Approach to Unpack Writers, Characters, and Personality in Historical Context

We present a new method for personality assessment at a distance to uncover personality structure in historical texts. We focus on how two 19th century authors understood and described human personality; we apply a new bottom-up computational approach to extract personality dimensions used by Jane Austen and Charles Dickens to describe fictional characters in 21 novels. We matched personality descriptions using three person-description dictionaries (Allport & Odbert, 1936; Goldberg, 1982; Saucier, 1997) and used Goldberg’s (1992) marker scales as reference points for interpretation. Factor structures did not show strong convergence with the contemporary Big Five model. Jane Austen described characters in terms of social and emotional richness with greater nuances, but using a less extensive vocabulary. Charles Dickens, in contrast, used a rich and diverse personality vocabulary, but those descriptions centered around more restricted dimensions of power and dominance. Although we could identify conceptually similar factors across the two authors, analyses of the overlapping vocabulary between the two authors suggested only moderate convergence. We discuss the utility and potential of automated text analysis and the lexical hypothesis to (a) provide insights into implicit personality models in historical texts, and (b) bridge the divide between idiographic and nomothetic perspectives