Abstract
In the era of artificial intelligence (AI), graduate education in science, technology, engineering, and mathematics (STEM) demands innovative pedagogies that exploit human-AI symbiosis to foster learning without repressing human agency. This thesis explores the co-creation of deep learning (DL) teaching and learning materials by a student, professor, and large language models (LLMs). Studies include an exploration of LLM limitations, such as superficial responses and over reliance risks, as well as mitigation strategies to promote ethical, active engagement. Drawing on foundational texts like Deep Learning by Goodfellow et al. and Foundations of Deep Learning by Bishop and Bishop, the proposed approach integrates AI-assisted curriculum design, yielding a comprehensive module: PowerPoint slides on DL fundamentals (definitions, history, applications, regression/ classification examples), hands-on labs (polynomial curve fitting and MNIST digit recognition via feedforward networks), and assessments (iteratively refined homework, pre/posttests). Prompt engineering techniques ensured precise LLM responses, emphasizing clarity, action verbs, and iterative refinement. Deployed in University of Massachusetts Dartmouth courses (CIS 550 Advanced Machine Learning, Summer/Fall 2025; ECE 549/489 Network Security, Fall 2025), the module demonstrated efficacy through mixed methods evaluation. Pretest scores averaged 6.17 across courses, reflecting moderate baselines, while post test scores averaged 7.11. Lab results confirmed practical mastery, with 95% digit prediction accuracy in baseline network configurations and insights into overfitting and hyperparameter effects. Post-module surveys yielded high endorsement (M=4.89 on Likert scale items; 80-100% agreement on clarity, applicability, and confidence gains). Qualitative feedback praised visuals and theory-practice integration and offered suggestions for module enhancements like interactive elements. This model for human-AI collaboration validates scalable, equitable DL education. Future work includes larger-scale testing, ethical AI policy integration, and module enhancements, ensuring graduate student success as they assume partnership with LLMs in the age of AI.