Teaching with MATLAB Copilot

About MATLAB Copilot

MATLAB Copilot helps educators, students, and researchers create, refine, and debug MATLAB code with generative AI. It is used as a coding assistant, a tool that has gained popularity in the industry in recent years because it automates tedious tasks and improves productivity.

Embracing generative AI is key for developing the next generation of engineers and scientists. MathWorks includes access to MATLAB Copilot in the Campus-Wide License to help educators support this transition.

Opportunities for Generative AI in the Classroom

The use of generative AI tools requires students to learn additional skills such as creating prompts, refining prompts, iterating on ideas, and assessing the generated results. Each of these provides an opportunity to teach students new skills that they can use later in their careers.

Prompting: What are the best practices for creating prompts? How is prompting different from traditional search? What context can you provide to improve the prompt?

Refining and iterating: Did the generated results match your expectations? How can you get improved results from the system?

Assessing: Should you trust the generated results? How do you improve your confidence in the results? How can you test the results for correctness?

By teaching students how to use generative AI tools properly, educators can establish a foundation to accomplish more in a course than previously possible. Educators can spend less time troubleshooting code, leaving more time available to implement course objectives through hands-on projects.

Generative AI tools, like MATLAB Copilot, also provide real-time feedback and assistance to students so they can get help when they need it.

MATLAB Copilot recommends appropriate tools and functions, creating customized examples to help students identify an approach to a problem.

Students get explanations of MATLAB code or error messages, helping them to identify issues and continue with their work.

Students focus more on the domain-specific nature of their work, delegating implementation details to MATLAB Copilot.

An example of a course that teaches such skills is AI-assisted MATLAB Programming with ChatGPT.

Concerns about Generative AI in the Classroom

While generative AI tools bring many benefits, they can be disruptive when assessing student skills. This can be true in foundational courses where fundamental concepts are assessed with homework problems that are now trivial for a modern generative AI tool. In this case, instructors may wish students to perform exercises without generative AI assistance (see the Configuring MATLAB Copilot section below).

Knowing that generative AI tools are here to stay, instructors are exploring and implementing new assessments that prioritize student understanding over programming language syntax. Examples of this include focusing on:

Project-based learning
Critical thinking
Systems-level thinking

Research on using generative AI tools for assessing student work underscores the potential risks regarding bias, validity, and transparency. MATLAB Copilot is not intended for student assessment. Instead, MathWorks offers MATLAB Grader™—a browser-based platform for creating, sharing, and automatically grading formative and summative assessments while providing immediate feedback on student progress. MATLAB Grader is deterministic and does not use AI to evaluate learners’ work. Because MATLAB Grader is designed to measure students’ ability to solve problems independently, MATLAB Copilot is disabled when students complete MATLAB Grader assessments.

Configuring MATLAB Copilot

While MATLAB Copilot excels at introductory programming tasks, it is still crucial for students to build a solid foundation in programming independently so they can address more advanced tasks in the future. MATLAB Copilot settings can be used to control the functionality of MATLAB Copilot (enable/disable autocompletions or enable/disable MATLAB Copilot entirely).