Augusta details open-source plans for an AI tutor that coaches coding students

Augusta University has detailed plans to make SE Tutor, an AI-supported software engineering tutor, open source, with free training materials and workshop recordings. In a November 4 account, the university also set out a goal of reaching more than 2,000 students across a three-university collaboration. The tool is being designed to coach students through problems rather than supply complete solutions.

The development project brings together Augusta, the University of Alabama and Willamette University through a five-year, $2 million U.S. National Science Foundation grant. Willamette first described the award on September 4 and is leading the collaboration. Augusta’s new account adds detail about the intended reach and distribution of the platform; it is not a release of finished software.

Turning a warning into a question

SE Tutor’s starting point is the feedback students already encounter: failed tests, security alerts and code-quality reports. Its planned role is to use generative AI to translate that technical feedback into guided questions. Instead of stopping at the location of an error, those questions are intended to help students understand its cause and reason through a correction.

Augusta gives a security example. If a development tool flags a potential problem with user input, the tutor might ask what could happen when someone enters unexpected information. It could then prompt the student to consider how to test whether the program handles that input safely. The student would investigate, change the software and evaluate the result.

That design deliberately keeps the reasoning with the learner. Augusta contrasts it with a general-purpose chatbot that can hand over a complete solution. The project’s stated goal goes beyond fixing the current program: students should become better able to evaluate their own work and apply what they learn to later problems.

When students encounter a problem in their code, we want them to understand why it happened and how to approach similar problems in the future.

Gursimran Walia, Augusta University principal investigator

A testing tutor provides the starting point

The researchers previously developed Testing Tutor, another NSF-funded tool. Willamette says empirical studies found lasting improvements in students’ software-testing skills from that work. The earlier system built question-driven feedback around the output of professional development tools, helping students write more effective tests and higher-quality software, according to the university.

SE Tutor is intended to extend that approach beyond testing to the full software development process. Educators would be able to connect professional tools and turn their dense output into teaching feedback. The team also plans to investigate additional learning theories. The reported results belong to Testing Tutor, not to the new platform still under development.

Building the classroom support alongside the software

The participation target includes more than 50 faculty members across the three universities. Workshops on each campus are intended to help instructors connect SE Tutor with tools they already use and create assignments that support and assess learning. Undergraduate and graduate students will also have research opportunities to help design, develop and study the platform.

The open-source plan would give educators beyond those campuses access to the software and free supporting materials. The team also plans a way for instructors to share resources and examples of how they use SE Tutor in courses. That makes teaching support part of the planned distribution, rather than leaving other institutions with the software alone.