Natural language processing word dictionaries for personnel selection: Achieving explainable, generalizable, and incremental prediction with low impact when passively scoring narrative application information

The purpose of this research is to show how natural language processing word dictionaries can measure job-related attributes for personnel selection. Word dictionaries are lists of words and phrases that reflect a construct and can automate the scoring of narrative application information (e.g., interview responses, essays, descriptions of past jobs), even if the information is passively scored (i.e., data collected for other purposes in the selection process). We develop 17 dictionaries that represent competency-, personality-, and motivation-based language. We apply these dictionaries to passively score narrative application information to predict selection ratings and decisions and postselection performance in 12 samples from six studies (total N = 67,652), including employment and school settings, experimental and operational contexts, concurrent and predictive designs, and a range of occupations. Results show consistent prediction of selection ratings and decisions and postselection performance with uncorrected grand averages across dictionaries and studies in the low .20 s for selection ratings and nearer .10 for performance criteria, incremental prediction beyond traditional assessments (mental ability and personality tests), measurement of constructs like motivation and personality without using self-scoring, and small gender and racioethnic subgroup differences. We close with theoretical and practical implications to show that dictionaries are explainable, generalizable, straightforward to content validate, useful for passively scoring information already gathered, accessible to researchers with typical statistical skills, can add incremental prediction to common existing measures, can capture previously difficult-to-measure constructs like motivation, and may help address the validity-adverse impact trade-off. (PsycInfo Database Record (c) 2026 APA, all rights reserved).