Event Actions
Computing Education in the Era of Generative AI
Abstract:
The advent of generative AI (GenAI) is already changing the landscape of computing education. Researchers were quick to show how GenAI could solve most introductory programming problems and score well on exams compared to students. Traditional courses in the introductory sequence have placed quite a bit of weight on student ability to write correct solutions to programming assignments, while the pandemic moved many assessments and assignments into the virtual space. The probabilistic nature of GenAI makes it difficult to find students using it in inappropriate ways; every student in a class could use it and submit different programs. Making matters worse, current research is exposing learner over-reliance on these models, showing that usage of GenAI can inhibit the development of problem solving and computational thinking. One recent study that asked students to solve a programming problem with GenAI coding tools found students with high self efficacy could accelerate to a solution, while students with low self efficacy were stymied, misled, and confused by GenAI.However, there are rays of hope. Current models can provide correct and detailed feedback on programming assignment submissions, help students understand cryptic error messages, and act as a virtual TA available at all hours. Tools like these could extend help to students who are afraid to ask or are deterred by long lines to speak to the professor or TA, potentially broadening participation in computing. Current research has revealed the possibilities to scale grading of tasks using GenAI that we know are helpful to students but difficult to grade, such as "Explain in Plain English." New types of pedagogical assessment that utilize GenAI are also emerging, such as Prompt Problems, where learners iterate on prompts to generate code that can pass all test cases. The possibilities already uncovered hint that we are on the cusp of revolutionizing computing education. In this talk, I will present a current look at what we know from user studies on how students use GenAI while learning programming and what these insights might mean for the future of the field.
About the Speaker:
James Prather is an Associate Professor of Computer Science at Abilene Christian University. He holds a masters degree in neuroscience, two masters degrees is theology, and a Ph.D. in Computer Science. His primary research area is in Human-Computer Interaction in the domain of Computer Science Education. Over the last five years he has won five best paper awards for his research on novice programmer metacognition, interaction with and design of error messages, and user interaction with generative AI. James is currently a Program Co-Chair of the SIGCSE Technical Symposium, the largest computing education research conference in the world, on the program committees for several other conferences in various roles, and on the advisory board for GitHub Education. His extensive consulting experience includes research for Google and UX research for iHeartRadio.James lives in Abilene, TX with his wife and four kids. He enjoys traveling with his family, camping, and reading sci-fi and fantasy novels, sometimes all at the same time.