Research-Led Learning in Action: Classifying Hope in Political Texts Through Student Involvement

This paper investigates the classification of hope in political texts, presenting findings from a cross-disciplinary course offered at an Irish university. Students became involved in the research process during a lecture by manually identifying expressions of hope in a political speech. The paper examines the levels of agreement and disagreement among the students before testing various computational methods for measuring hope in texts. These methods include dictionary-based classification, machine learning classifiers, state-of-the-art transformers, and advanced large language models like GPT-4. The analysis reveals significant variation in students’ perceptions of hope and highlights the challenges that automated classifiers face in aligning with human judgement when identifying expressions of hope. The findings not only enrich our understanding of political discourse surrounding hope but also illustrate ways to integrate research-led learning into lectures.