APPLICATION OF DEEP LEARNING APPROACHES TO IMPROVE STUDENT PERFORMANCES IN BIOLOGICAL SCIENCE EDUCATION AT FEDERAL UNIVERSITY OTUOKE

Abstract

,This study examined the effectiveness of deep-learning instructional approaches in improving undergraduate students’ performance in Biological Science Education at Federal University Otuoke. Using a quasi-experimental pretest–posttest control-group design, 83 second-year Biology Education students were randomly assigned to experimental (deep-learning) and control (traditional instruction) groups. The intervention integrated a fine-tuned ChatGPT model, a BERT-based real-time feedback system, and a learning analytics dashboard. Results showed that students exposed to deep-learning approaches achieved substantially higher post-test scores than those taught through lecture-based methods, indicating the superior efficacy of adaptive AI-supported instruction. While both male and female students improved, descriptive statistics revealed greater gains for females, whereas inferential analysis indicated a statistically significant difference favouring males with a very large effect size. These findings underscore the transformative potential of deep-learning tools in Nigerian higher education while highlighting gender-related complexities. Implications, limitations, and directions for future research on equitable digital pedagogy were discussed.