Report Urges Colleges to Rethink Long Math Prerequisites Before AI Courses

A different path into AI coursework

In many traditional computer science programs, students may need to complete four or five dedicated mathematics courses before they can enter core AI classes. The typical sequence includes calculus, discrete mathematics, probability and statistics, and linear algebra. As colleges and universities build new undergraduate AI pathways, the CRA asks whether all of that mathematics really has to come first.

The NSF LEVEL UP AI workshop recommendation report Math for AI answers with a different approach. According to the CRA, it is designed to reduce mathematics prerequisite bottlenecks without reducing the mathematical sophistication students need to understand AI. The report came out of workshops that brought together 22 participants from 21 institutions. Its authors are Michael Peterson (University of Hawaii at Hilo), Isaac Shaffer (Northern Arizona University), Enrico Pontelli (New Mexico State University) and Noah Cowit (Computing Research Association).

Modules instead of a long course sequence

Rather than requiring a series of traditional math courses up front, the report recommends organizing the mathematical content for AI into more focused modules. These could be taught in dedicated "Math for AI" courses or built directly into computing and AI courses. One example given is placing statistics content inside an Introduction to Data Science course, so students meet the mathematics alongside the computing concepts and applications that make it relevant.

The report identifies learning areas in discrete mathematics, calculus, probability and statistics, and linear algebra, plus prerequisite topics and desirable electives. Specific concepts range from logic, combinatorics, derivatives and gradients to probability theory, model evaluation, vector spaces, and eigenvalues and eigenvectors.

It does not prescribe exactly when each topic should be taught. Because AI pathways have different objectives, the report recommends that programs decide which concepts are foundational and should appear repeatedly, which need deeper treatment at some point, and which require only familiarity later in a program. What changes is the assumption that all of this learning must happen before students can engage meaningfully with AI.

Peterson, Chair and Associate Professor of Computer Science at the University of Hawaiʻi at Hilo, said: "The goal isn’t to take mathematics out of AI education, it’s to be much more intentional about what mathematics students need and when they need it."

Real-world examples and cooperation with math departments

The report also recommends tying mathematics to applied, real-world AI examples. The CRA says this can help students understand implementation and make clearer why particular concepts matter for their AI learning and future careers. Pontelli, Regents Professor of Computer Science at New Mexico State University, said that integrating the two "gives us an opportunity to make mathematics more relevant to students while still building the foundation they need to understand AI deeply."

The report further encourages computing programs developing "Math for AI" courses to work directly with their institutions' mathematics departments, including potentially creating courses co-taught by computing and mathematics faculty.

What it means for institutions

The report's framework is meant to give institutions a common starting point for thinking about the mathematical preparation AI students need, instead of each program developing its own approach independently. Shaffer, a Lecturer at Northern Arizona University, said a coordinated approach "can make high-quality AI education more achievable for institutions that simply don’t have the resources to build every component themselves."

For institutions designing new undergraduate AI programs, the CRA says the recommendations offer a chance to plan the mathematics and AI pathways together from the start. Existing programs are given a reason to examine whether their prerequisite structures help students develop the math they need or unnecessarily delay their entry into AI. The CRA frames the question as how to provide the right mathematics at the right time, without creating unnecessary barriers.

Math for AI is one of seven NSF LEVEL UP AI workshop recommendation reports.