Half of 6,065 Arizona middle schoolers solved at least two problems a week on an AI math platform. They grew 0.11 SD more on MAP than similar peers. At 20+ a week: 0.31.
Observational, and it measures platform use, not the tutor alone.
Ask every vendor what share of kids hit the dose
1
10
Alam, Dominguez-Mel and Liu, EdWorkingPaper 26-1611, Oct 2026 (observational, propensity-weighted). Hechinger coverage Nov 2: hechingerreport.org/proof-po…
doi.org/10.26300/vcdh-yc23
6
Same three practice sets: all in one day, or one a day for three days.
Six weeks later, the spaced 7th graders did better.
Spread the practice out
1
26
Barzagar Nazari and Ebersbach, 'Distributing mathematical practice of third and seventh graders', Applied Cognitive Psychology 2019 (213 students, in school)
eric.ed.gov/?id=EJ1262987
14
Researchers randomly assigned four math curricula across 110 elementary schools in 12 districts.
In 2nd grade, Saxon students outscored Scott Foresman-Addison Wesley students by 0.17 standard deviations, the only gap that survived a multiple-comparisons correction.
Curriculum choice moves scores. Ask your publisher for a head-to-head trial.
1
31
Agodini and Harris (Mathematica), randomized trial of four early elementary math curricula, IES/NCEE 2010-11; summary paper ERIC ED528687
files.eric.ed.gov/fulltext/E…
18
Kindergartners scoring in the top 5% in math: by 4th grade, 41% of the higher-income ones were still there. Lower-income: 26%.
From 4th to 8th grade, it's 80% vs 72%.
The gap is widest in K-4. Check every kid every year, and keep advanced work open to the ones who slip.
1
25
NWEA research brief, 'Missing at the top: Lower-income students underrepresented among high achievers', Nov 2026 (MAP Growth math, nearly 1 million students, 2021-22 to 2025-26)
nwea.org/uploads/Missing-at-…
19
Clement, Lochhead and Monk, 'Translation Difficulties in Learning Mathematics', American Mathematical Monthly, April 1981
doi.org/10.1080/00029890.198…
23
Swap "find angle ABC" for "find as many angles as you can."
In John Sweller's studies, students given the open version learned more.
Try it on your next warm-up
1
33
Sweller, Ayres and Kalyuga, 'The Goal-Free Effect', Cognitive Load Theory (Springer, 2011), ch. 7; original studies include Sweller, Mawer and Ward (1983)
link.springer.com/chapter/10…
20
783 tutors were randomized. Half got AI suggestions for what to say after a student's mistake.
Students passed the exit ticket, which tutors can't help with, 66% vs 62%. With lower-rated tutors: 65% vs 56%. No significant year-end gain.
Give it to your weakest tutors first.
1
28
Source: Wang, Ribeiro, Robinson, Loeb and Demszky, Tutor CoPilot (RCT, 783 tutors, 1,013 students, grades 3 to 6)
arxiv.org/abs/2410.03017
22
Across 7.5 recorded hours with their toddlers, one parent said 4 number words. Another said 257.
At 46 months, that gap showed up in which kids understood what number words mean, even with family SES controlled.
Talk numbers at home
1
32
Source: Levine, Suriyakham, Rowe, Huttenlocher and Gunderson, Developmental Psychology, 2010. 44 children, five 90-minute home visits from 14 to 30 months. Correlational.
pmc.ncbi.nlm.nih.gov/article…
23
Tina Akers-Porter, a Modesto Junior College math instructor, called the extension "better than nothing."
California's community colleges now have until July 2028 to prove precalculus helps STEM majors. None has yet.
The state says its lowest-placed students more often finish calculus in year one by starting in it.
Show that data by college
1
50
Source: EdSource, Nov 12, 2026, and the CA Community Colleges Chancellor's Office memo ESS 26-67 (Oct 13, 2026)
edsource.org/2026/california…
20
Coaching averaged 0.18 SD on achievement in 60 causal studies. Big programs got a fraction of small trials' effects.
Dylan Wiliam's slides show PD tested at scale: randomized, 22,709 students in England, 0.13 SD on exams, $1.50 per student a year.
Judge PD by what scales.
Earlier today, I gave a presentation at @NCME38's Classroom Assessment conference in Minneapolis titled "Effective, scalable, sustainable professional development on formative assessment: A 30-year retrospective". The slides are here: bit.ly/3Utpz1W
1
37
Source: Kraft, Blazar and Hogan, Review of Educational Research, 2018. 60 causal studies, mostly US early-grade literacy. Trial figures are the EEF evaluation (Speckesser et al. 2018) in Wiliam's slides.
doi.org/10.3102/003465431875…
17
Same lesson on 4+3+6=__+6. Same gains right after it.
Four weeks later, third and fourth graders told to gesture the strategy kept 85% of their gains. Kids told to say it without gesturing kept 33%.
Tomorrow, let them move their hands.
1
32
Source: Cook, Mitchell and Goldin-Meadow, Cognition, 2008. 84 third and fourth graders, mathematical equivalence, four-week follow-up.
pmc.ncbi.nlm.nih.gov/article…
27


