THE CREATIVE FORCE
Harvard's Center for Digital Thriving named this week's problem: "AI is making student learning invisible to teachers." Offering Maryland undergraduates an AI tutor pulled them away from their own courses. Our defenses shrink the work further; as Jennifer Taback writes, "we have narrowed assessment to what we can proctor." But mathematics is a creative act you can only judge if you can see it: a question posed and lived with, the way young mathematicians took four papers to crack Graham's 1971 conjecture. So by June, where will the math we grade actually happen?
THE FIVE
- [Research] EdWorkingPapers: The Effects of Course-Integrated AI Tutoring on Student Performance and Engagement: A Randomized University Trial In a University of Maryland randomized trial, sections given a university-provided tutor built on OpenAI's GPT-4o earned final grades 0.37 standard deviations below other sections of the same course, though only about 15% of students ever used it (working paper, not yet peer-reviewed). Before any AI rollout in math, ask what it will replace.
- [Report] Harvard's Center for Digital Thriving: An AI Policy Isn't a Playbook Among teachers who had faced an AI dilemma, 74% described something about cheating, and the side effect is perverse: "Some students deliberately underperform to avoid being flagged." In math, suspicion teaches students to hide exactly the thinking we're trying to grow.
- [Essay] Terence Tao's blog: What does the advent of powerful AI models mean for mathematicians like me? Bowdoin's Jennifer Taback resents AI for pushing her assessments into timed classrooms, since without graduate student help, oral exams are "impractical at the scale of my teaching." Yet AI is also the research colleague her small department can't provide: a double bind teaching-heavy departments will recognize.
- [Policy] CalMatters: Student math scores are abysmal. A new California law requires early testing for skill SB 1067 will screen every California K–2 student on counting, quantities, and number recognition starting in 2028, with no new money attached. The state's largest teachers union opposed it, arguing teachers already know who is struggling; the real test is whether screening changes what happens next.
- [News] Quanta: Mathematicians Harness Randomness To Crack a 55-Year-Old Conjecture Ronald Graham asked in 1971 whether any set of nonzero numbers on a prime-hour clock can be ordered so every running total is different; four papers by young mathematicians now prove it for every large enough prime. Noah Kravitz's motive, "it's embarrassing for humanity that we don't know this," is the creative force no proctored test measures.
THE POLL QUESTION THIS WEEK
By June 2027, where will most graded math work happen in your classes? We'll check back then. Vote below, then tell us about one assignment you've already moved this year and what it cost. Can't pick? Tell us what it depends on. (single choice below).