Character Development in Oral Testimonies: Computational Modeling of Religiosity in Holocaust Narratives

Character is a core feature of oral testimonies, inviting readers into the protagonist’s process of identity formation. This work explores character trajectories in 1,000 Holocaust testimonies from a religious development perspective, focusing on religious practice and belief. We extract religious trajectories by fine-tuning large language models and then perform a quantitative analysis across the narratives. The trajectories are clustered to identify common structures of religious evolution, yielding a taxonomy of five structural types. The analysis is validated automatically, manually, and through close reading of selected examples. We find that the most common structure for religious practice involves oscillations, whereas a constant-positive pattern is most prevalent for religious belief. Finally, we demonstrate how the proposed method can be used to select similar testimonies. While some results support existing theories, others offer new insights into the complexity of survivors’ lived religious experiences, highlighting the potential of computational methods for testimony analysis.