Exploring students’ emotions towards programming: Analysing sentiments using concurrent conversion mixed methods


ATMAN USLU N., Onan A.

Education and Information Technologies, 2025 (SSCI) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1007/s10639-025-13482-z
  • Dergi Adı: Education and Information Technologies
  • Derginin Tarandığı İndeksler: Social Sciences Citation Index (SSCI), Scopus, Communication Abstracts, EBSCO Education Source, Educational research abstracts (ERA), ERIC (Education Resources Information Center), INSPEC
  • Anahtar Kelimeler: Anxiety, Boredom, Enjoyment, Gender differences, Hope, Programming education
  • Dokuz Eylül Üniversitesi Adresli: Evet

Özet

Understanding the emotions experienced by programming students, particularly concerning gender and education level, is increasingly critical. However, only limited research has used text data to examine these differences within the context of programming education and emotions. This study aims to determine programming students’ emotions and any differences based on gender in secondary and higher education and compare the performances of the algorithms used in prediction with sentiment analysis. The study uses concurrent conversion mixed methods and data from two study groups. The first group consisted of 444 secondary school students who completed an electronic questionnaire created for this study. The second group comprised 202 first-year software engineering and computer science students. The results of independent sample t-tests revealed significant differences in enjoyment, anxiety, boredom, and hope scores among secondary school students based on gender. The t-values for each category were as follows: enjoyment (t = 2.333, p <.05), anxiety (t = 2.519, p <.05), boredom (t = 3.841, p <.01), and hope (t = -3.829, p <.01). Among middle school students, girls reported higher scores in enjoyment, anxiety, and boredom compared to boys, while their hope scores were lower. However, no statistical differences occurred between females and males at higher education levels. Sentiment analysis revealed that BERTurk algorithms achieved higher accuracy scores than machine learning. BERT produced 96% accuracy for enjoyment, 92% for hope, 97% for anxiety, and 96% for boredom, while support vector machines and random forest algorithms achieved 94% accuracy in predicting positive and negative emotions.