What do our core school subjects – including math – look like in the age of AI? With today’s tools, the entirety of human knowledge is immediately accessible, computational procedures of any scale are not only possible but also cheap to automate, and pocket-size digital devices are so ubiquitous that it’s rare to catch a human without one.

In short: breadth, depth, and immediate access to most of what we teach in our existing curriculum, at any time and place.

This new reality is profound, and might imply an extreme but logical solution: no need for math, science, or most of our other subjects as-taught today. It’s wasted time, right?

Yet our education system has a strong historic emphasis on imparting math. Mathematics is our most ancient classroom subject, with the followers of Pythagoras recording and transferring early mathematical proofs to one another in the 6th Century BC. The term “mathema” literally translates to “subject of instruction” in Ancient Greek. From ancient civilization through the Sputnik era to the computer revolution, mathematics education has managed to resist nearly every wave of existential threat from technology that otherwise implied its automation. Hand-held calculators? We still teach long-division and times-tables. Digital graphing software? We still hand-write the algebra to solve y = mx+b. Any rethinking may risk defying some ancient wisdom, or at least represent a fool’s errand against an immovable tradition.

However, an outright defense of our approach to math education also assumes it’s working in the first place. 78% of high school graduates fall below NAEP proficient on the 12th grade “Nation’s Report Card” assessment, a trend that has continued with robust consistency for the past two decades. Our international ranking in math scores is a worn-out horse of a news story, from the 1980s to just last year. And the resulting levels of “math anxiety” – a literal fear of numbers or “being a math person” – is estimated to impact as many as 93% of U.S. adults. At that scale, our national math problem necessarily impacts both our G.D.P. and our civics in dramatic yet still unmeasured ways.

The most alarming of all: the share of students reporting that “math will help me when not in school” has declined since 2017 – suggesting a hidden “relevancy crisis” lurking beneath the surface. The question “why am I learning this?” is every math teacher’s dreaded quagmire today. In a world in which students grow up with digital devices that automate so much of the math curriculum, we should expect this to only increase.

A holistic conversation on numeracy in the age of AI needs each of these three starting points together: acknowledging 1) present technology can automate most of the mathematics we currently require, that 2) our system as-is fails to empower the majority of students in mathematics anyway, and yet 3) even when presented with prior automation, we have still chosen again and again to teach some manual procedures, because there may be value for muscle-memory that is “built-in.” Especially given how often we need math in daily life situations and to simply question ideas, it seems dangerous to abandon them completely.

So now what?

Most historic debates on math education only focus on one of these dimensions and conveniently ignore the others. We need a better framework for the problem that recognizes all of these realities simultaneously and integrates them. Here’s a few approaches (with active momentum behind them in the field) that may hit three birds with one abacus:

1. Conceptual re-leveling: in a world of AI, not every concept in the math curriculum demands automaticity gained through repeat practice and drills; some may be better served at a conceptual “supporting” or even “appreciation” level, which has been argued by Phil Daro – one of the original Common Core authors. Most math standards provide guidance on what to teach, but not at what depth or for how long. In the status quo, this results in students learning many concepts a mile-wide and an inch-deep, with the assumption that all concepts are created equal, limiting time for project-based learning, durable skills transfer, or applying math concepts in the context of real-world data and phenomena. Instead, we can rebalance time away from memorizing many formulas towards deeper fluency in the most critical topics, while better emphasizing both proof-based argumentation and digital daily life applications in statistics and data analysis. Contact Student Achievement Partners or Data Science 4 Everyone for more information on the project, including on pilots in four states.
2. Interest-based math pathways: this approach transitions high school math from a “one-size-fits-all” approach towards the autonomy offered in a typical college experience, where students can choose between the math content that best fits their interests and career aspirations. STEM or Economics students may pursue Calculus, Psychology or Law students may pursue Data Science & Statistics, and Humanities or Arts students may pursue Quantitative Reasoning courses in high school. “Why am I learning this?” becomes a question students get the chance to answer themselves. See the U.T. Austin Dana Center Launch Years initiative and Data Science 4 Everyone (our team), who have engaged over 25 states in developing this model.
3. Math-badging: intense debates have permeated the grade-level timing of courses like Algebra 1 and the existence of “honors” vs “non-honors” math placement. Yet these arbitrary distinctions could vanish in a system of self-paced learning that still recognizes the pre-requisite model of math education with module-based assessments. Systems that enable self-paced learning could support this content solution in what otherwise is historically perceived as the most rigid school subject. See the XQ Institute for their pilot in three states.
4. Non-linear progressions: the present-day math curriculum assumes mathematics “stacks” in a linear trajectory (e.g. you must learn basic Algebra before Calculus, or addition before multiplication). Yet learning sciences literature at both the micro- and macro-level is questioning that assumption, where students have been observed to learn more effectively in both interleaved formats (mixing problem types to encourage strategic planning for a problem up-front, drawing on multiple areas of knowledge) or through integrated mathematics (learn many concepts at once and the connections between, rather than one siloed course at a time).

Our current inflection point invites new approaches that can simultaneously increase relevance and confidence for students, recognize some baseline quantitative reasoning and fluency is critical for everyone, and as Conrad Wolfram argues, “teach math as if computers exist.” Our historic failures in math education should not be a national mourning, but rather a license for creativity unlocked by technology. If we get this right, the impact may be infinitely larger than we could imagine.