Automated discourse analysis for attitudinal profiling in textual data

Abstract

Competency-based learning is a transformative approach that seeks to integrate knowledge, skills, and attitudes (KSA) (Zabala & Arnau, 2015) in the development of learners. In this context, attitudes—understood as observable behavioral tendencies shaped by affective, cognitive, and conative components—play a crucial role in shaping professional identity and decision-making. In military education, the Brazilian Army’s NDACA framework formalizes strategies for attitudinal development and evaluation through structured pedagogical practices. Despite the growing interest in using Natural Language Processing (NLP) to identify behavioral traits in text, few studies focus on attitudinal profiling through discourse analysis. To address this gap, we developed a model that leverages Large Language Models (LLMs) to infer and classify attitudinal content from open-ended textual responses. Applied to responses from 14 military students enrolled in a "Leadership and Management" course, the model demonstrated promising results in detecting patterns aligned with the NDACA framework. These findings suggest that LLM-based methods may support attitudinal assessment in educational contexts (Henklein & Carmo, 2013), offering scalable and cost-effective insights into learners' values, dispositions, and behavioral trends.