Speaker: Joshua Jackson

Institution: Washington University in St. Louis

Date: 14:00 BST, 17th September, 2026

Talk Title: LifeSentence: Language models can encode human life course trajectories from longitudinal panel data

Abstract: LifeSentence adapts a pretrained large language model to reason over longitudinal panel data by representing life events as structured natural-language records. Trained on approximately 65,000 participants from the German Socio-Economic Panel, the model is evaluated across prediction, robustness, and reasoning tasks. It outperforms classical and deep-learning baselines, reconstructs event order without timestamps, and recovers established patterns such as the education premium, gender wage gap, and motherhood penalty without explicit supervision. The paper argues that pretrained distributional knowledge can make life-course modelling practical with smaller panel datasets while supporting new predictive and counterfactual questions about human biographies.

Bio: Joshua Jackson is the Saul and Louise Rosenzweig Professor of Personality Science and a professor of Psychological & Brain Sciences at Washington University in St. Louis. He studies personality development and assessment, including the genetic and environmental factors associated with personality change, the role of educational experiences, and how different assessment methods shape estimates of personality development. He received his PhD from the University of Illinois Urbana-Champaign and his BS from the University of Wisconsin–Madison.