When I was involved with organizing public events at Brookings, the main problem was audience members giving speeches instead of asking questions. They would come across as wannabe guest speakers.

Delighted that 'Embedding formative assessment'—which I wrote with Siobhan Leahy—is now available from Solution Tree in both paperback and e-book format: bit.ly/3KeJp8c

Prediction: systematic reviews and meta-analyses are some of the things that AI will very soon be doing better and faster than us mere mortals

Large language models score close to, and sometimes out-perform, humans on tests of 'theory of mind' (tracking the mental states of others): go.nature.com/3UNNqFm

Outgoing (but not shy) teachers regard more talkative children as more intelligent (HT InnerDrive): bit.ly/3V1MkX0 ($)

In 2011, Anjum Halai and I edited a collection of papers exploring whether research methods of the 'North' are appropriate for doing educational research in the global 'South'. I am delighted to announce that this collection is now available as an ebook: bit.ly/3WVLkpW

This is how you write a title and an abstract. // The University of California Was Wrong to Abolish the SAT: Admissions When Affirmative Action Was Banned onlinelibrary.wiley.com/doi/full/10.11…

If your child’s school cuts social studies and science to make more time for reading instruction, it’s a good indication they don’t understand how reading comprehension works.

Virtual tutoring in grades K-2 boosts students' early literacy skills, especially for students who were behind, and especially for 1:1 tutoring edworkingpapers.com/ai24-955

A policy limiting K-2 suspensions 1) reduced suspensions, 2) substantially closed the Black-White suspension gap, 3) increased achievement for high-risk students, and 4) had no negative effects on achievement or discipline for other students. edworkingpapers.com/ai24-950

OK; this is definitely the nerdiest thing I've ever asked on Twitter/X. Is it more helpful to regard construct under-representation as a property of tests or of test scores? Most people seem to assume the former but the lack of symmetry with construct-irrelevant variance bugs me.

Dylan Wiliam 𝐁𝐚𝐫𝐫𝐲 𝐆𝐚𝐫𝐞𝐥𝐢𝐜𝐤 Robert Craigen Natalie (Mullen) Brankin Anna Stokke Alt B This is what he's referring to. csun.edu/~vcmth00m/long…

A large language model optimized for diagnostic reasoning produced better differential diagnoses (DDx) for 302 challenging real-world medical cases (sourced from NEJM) than unassisted clinicians, and even better than clinicians supported with AI: bit.ly/4dNmyOz

Assessment serves as the bridge between teaching and learning. 🌟

Join Dylan Wiliam on June 3rd to unpack the methods behind assessment and to learn how educators can enhance decision-making and improve student learning outcomes.

Register now: ow.ly/GBst50RotrC

Student course evaluations are biased and lack validity for use in any consequential decision.

But OTOH my course evaluations are excellent because I am a very good teacher.

Given the small number of people who use frontier models today, I think many people who found AI useless when using the free version of ChatGPT are going to be shocked by the capabilities when GPT-4o is made freely available worldwide.

Expect a burst of concern & excitement.

All students benefit from attending effective high schools but the least advantaged benefit more in terms of HS graduation, college-going, & arrests in school: bit.ly/4alzwjK ($). Also, test-score based value-added measures underestimate benefits for the least advantaged

All students benefit by attending more effective high schools but the least advantaged gain more in HS graduation, college enrollment & school-based arrests: bit.ly/4alzwjK ($). Also, test-score based value-added measures underestimate benefits for disadvantaged students

I bristle whenever a skeptic is dismissed as 'cynical.' There's a huge difference. Skeptics think and doubt and, in so doing, affirm a vision of the way things ought to be. Cynics affirm nothing, take no action, and end up helping to perpetuate an objectionable status quo.

Is it true that some students just can’t do math? No, but you probably shouldn’t design instruction so that only a small minority of students can do it!

aft.org/sites/default/…