TEACHING SOFTWARE DEVELOPMENT IN THE AGE OF AI LLMS: WHY ABSTRACT MATH AND LOGIC MATTER MORE THAN EVER
M. Fonkam1, N. Vajjhala2
The explosion of freely and easily accessible content on the Internet already presented major challenges for academia on how best to leverage technology and online resources to improve student learning while mitigating the dangers. The recent introduction of AI models like ChatGPT, Deep Seek and related LLM models compounds these dangers. Most academics will agree that in today’s teaching and learning environment the objectives of the pedagogists (instructor and support team) do not always align with those of the learners. While the pedagogists emphasize the learning outcomes typically expressed as demonstrable knowledge, life skills and even learner attitudes, many learners are mostly focused on getting as good a passing grade as the means at their disposal will allow. Ideally a student’s grade from a course should reflect the level of achievement of the learning outcomes. The traditional approach to pedagogy, especially the component of grading and evaluation of student learning are seriously challenged by the ease with which learners can simply discover solutions online or create them using AI assistants. In this presentation, we explore the challenges, opportunities and strategies to teaching software development in the face of AI assistants such as Github Copilot, Deepseek and ChatGPT. We argue that with the growing complexity of software and the multi-paradigm nature of mainstream programming languages (Python, Java, JavaScript, etc) there is now, more than ever before, a need to repurpose the learning outcomes and revise the approach to pedagogy. We explore the growing imperative of abstract Mathematical, logical thinking and high-level problem framing skills for addressing the new challenges in modern software development. We then present a new set of learning outcomes and strategies to implementation including the use of pure Functional Programming and Logic Programming languages such as Haskell and Prolog respectively. As AI technologies become part and parcel of the Internet infrastructure, the onus falls on us in the teaching profession to embrace them in a way that enhances rather than compromises student learning. This demands a fundamental shift in the way we do business, engage with the learners and ultimately evaluate their learning. This work represents is a small step in that direction.
Keywords: Learning outcomes, knowledge, life skills, learner attitude, abstract thinking, problem modeling.