Narrative Agency and Affective Valence in Prospective Teachers’ Pedagogical Autobiographies: A Computational Text Analysis of Two Cohorts

Reflective practice is central to contemporary teacher education, and pedagogical autobiographies are widely used as tools for reflection; yet the structural properties of these narratives, and their relationship to the reflective process they are meant to support, remain under-examined. This study applies computational text analysis to 361 autobiographical accounts written by prospective future Italian teachers across two consecutive cohorts (2024–25; 2025–26). The corpus covers eight pedagogical dimensions and is analysed through two rule-based indicators, Student Agency Ratio (SAR) and Sentiment Valence, both validated against human coding on a subsample, using methods that account for the nesting of dimensions within participants. Results show strong dimension-level differences in both SAR and valence, with agency and valence profiles that replicate across the two cohorts, whereas lexical richness shows a cohort-level shift. Assessment stands apart from the other dimensions: compared with them, it combines markedly more negative valence (a large within-participant effect, dz = 2.31), more frequent negative or unresolved endings, lower lexical richness, and predominant anchoring in upper secondary school. By contrast, STEM and support-related memories are more often anchored in primary school and are generally more positive. These findings suggest that prospective teachers enter initial training with domain-specific narrative schemas that may shape how they interpret key areas of pedagogical practice, especially assessment. As the data come from a single institution, the findings should be read as exploratory and warrant replication with larger, multi-institution samples.