Viraj Kumar
Eminent Speaker
Short CV:
Title of Talk 1: Should Programming Pedagogy and Assessment Change in Response to Advances in Generative AI?
Synopsis: A core objective in most computing curricula is to develop the ability of students to write correct, maintainable, and efficient code for certain tasks. The complexity of these tasks gradually increases from Introductory to Advanced programming courses. With breakthroughs in Generative AI techniques based on Large-Language Models, tools such as ChatGPT, Codeium, CodeWhisperer, Copilot, Ghostwriter, and Tabnine can now solve non-trivial code-writing tasks better than most students. Many of these tools are freely available.
This talk examines the following question: Does the proliferation of such tools demand changes to pedagogy and assessment for programming courses? If no changes are made, students may use such tools indiscriminately for short-term benefits (good grades) while failing to develop their code-writing abilities. Further, as such tools are increasingly integrated into professional software development contexts, banning their usage entirely may leave students ill-prepared for their professional careers. Instead, we argue for thoughtfully adapting pedagogy and assessment for programming courses and then explicitly introducing students to such tools.
Title of Talk 2: Helping Students Develop the Ability to Critique AI-Generated Code
Synopsis: As Generative AI technology matures, tools such as GitHub Copilot, CodeWhisperer, Codeium, etc. are able to generate more accurate code in response to natural-language prompts. However, these tools still occasionally make errors, which can be extremely subtle.
In this talk, we argue that educators must explicitly develop the ability of students to critique code. We introduce Refute questions as an easy-to-evaluate form of question to develop this ability, and we discuss strategies for using them as both formative and summative assessment items.
Title of Talk 3: Implications of Disruptive Technologies on Curricula, Pedagogy, and Assessment
Synopsis: AI-powered tools such as ChatGPT have the potential to both enhance and inhibit learning. As noted in India's National Education Policy (NEP 2020), "Particular attention will need to be paid to emerging disruptive technologies that will necessarily transform the education system... Our present education system's inability to cope with these rapid and disruptive changes places us individually and nationally at a perilous disadvantage in an increasingly competitive world." This talk will be in two parts. First, we will explore the capabilities and limitations of Large Language Models (LLMs), which power tools such as ChatGPT. Second, we will discuss the implications of LLMs on what we teach, how we teach, and how we assess.
Viraj Kumar
Qualifications: PhD, 2007
Title: Visiting Professor
Affiliation: Indian Institute of Science, Bangalore
LinkedIn:
Twitter/X:
Facebook:
Instagram:
Email: [email protected]