Explainable AI for Psychological Profiling from Behavioral Data: An Application to Big Five Personality Predictions from Financial Transaction Records

Every step we take in the digital world leaves behind a record of our behavior; footprint. Research has suggested that algorithms can translate these footprints into accurate estimates psychological characteristics, including personality traits, mental health or intelligence. The mechanisms by which AI generates insights, however, often remain opaque. In this paper, show how Explainable (XAI) help domain experts and data subjects validate, question, improve models classify traits from footprints. We elaborate on two popular XAI methods (rule extraction counterfactual explanations) context Big Five predictions (traits facets) financial transactions (N = 6,408). First, demonstrate global rule sheds light spending patterns identified model as most predictive for personality, discuss rules be used to explain, model. Second, implement local individuals are assigned classes because their unique behavior, there exists positive link between model's prediction confidence number features contributed prediction. Our experiments highlight importance both methods. By better understanding work general well they derive an outcome particular person, promotes accountability impacts lives billions people around world.