Don’t pull any old personality taxonomy from the shelf: The performance of historical and sample derived taxonomies in extracting personality information from text

Substantial efforts have been made to develop comprehensive taxonomies of personality traits in many languages. Nevertheless, given that what is important and salient in individuals’ lived experience might be changing over time, this raises the question about the long-term usefulness of ‘off-the-shelf’ taxonomies developed decades ago. In the current study we used a bottom-up approach to create a large sample-specific taxonomy of personality terms. We subsequently examined the overlap and sensitivity of this taxonomy compared to an established trait taxonomy in the same language. Overall, we found that the two taxonomies only showed limited overlap with a pronounced divergence in emotionality (Neuroticism) and social aspects (Agreeableness) of personality. In addition to this, we found that while the personality assessment extracted from self-descriptions using the established taxonomy showed alignment with participants’ self-rated personality, especially Extraversion, Agreeableness, and Neuroticism, the sample-specific taxonomy showed a significantly greater alignment between text-based and self-rated personality. In summary, our current study highlights the need to extend our thinking about the psycholexical hypothesis, moving away from assumptions of time invariant language encoding to more explicitly recognizing temporal and sample-specific dynamics underpinning the expression and use of personality trait terms.