This week's reading keeps coming back to one question: is AI helping students think, or doing the thinking for them? There's new evidence on homework versus exam results, a practical framework for structuring AI tasks, and some useful caution about translation tools and AI-first schools.
A CEPR study of more than 26,000 Year 7–12 students found AI use lifted homework scores by 18% and cut completion time by 30%, while monthly exam scores dropped 20% within six months. Students at a Hobart college describe patchy classroom rules, and neuroscientists warn that skipping the struggle of working through ideas may affect how deep thinking develops.
Why it matters: It's clear evidence that polished homework doesn't guarantee learning, a strong case for designing tasks where students do the thinking before AI helps.
UNSW educational psychologist Andrew Martin argues that most student AI use is currently ad hoc and unguided. He proposes structuring it around Load Reduction Instruction: reduce difficulty early, scaffold, practise, give feedback, then move students towards independence.
Why it matters: It gives teachers a research-backed sequence for AI-supported tasks, so students use AI to support their thinking instead of replacing it.
Nearly 80% of Australian students used generative AI for study last year, yet a growing number of young people say they avoid it over concerns about their thinking, jobs, bias and the environment. Several say they feel pressured to use it anyway so they don't fall behind.
Why it matters: Students' mixed feelings are an opening for honest classroom discussion about when AI helps and when it doesn't, which is a core part of AI literacy.
Districts are using AI translation to reach families who speak other languages, but experts warn it can give a false impression of accuracy, especially with tone, idioms and less widely spoken languages. Free tools often come without data-protection agreements, and high-stakes documents still need human checking.
Why it matters: Many Australian schools communicate with families who speak dozens of languages, so this is a useful checklist for when AI translation is safe to use and when a human is needed.
Researchers from MIT, Brookings and Digital Promise urge schools to focus on today's harms, such as data privacy and cognitive offloading, rather than doomsday predictions. They recommend narrow, carefully vetted uses of AI over rushed adoption or blanket bans, and one district involved a student advisory board in writing its policy.
Why it matters: It's a calm framework for school AI policy that sits between hype and panic, and involving students in the decisions is an idea worth borrowing.
A mixed-methods study of 302 primary teachers in Türkiye found moderately positive views of AI, with teachers more willing to use it than experienced with it. Teachers mostly used ChatGPT, GeoGebra and Photomath for planning, visualising concepts and feedback, and they raised concerns about reliability, privacy and suitability for young children.
Why it matters: Primary-specific evidence is rare, and the finding that training and regular use most strongly shape teacher confidence is directly relevant to PD planning.
Benjamin Riley argues that Alpha School, the widely publicised AI-driven school model, rests on flawed ideas about how people learn. He points to tiny enrolments, very high fees and weak results at its Arizona charter school.
Why it matters: This is a useful counterweight when leadership or parents ask whether AI tutors could replace teachers: the evidence so far says relationships still matter.
A five-minute read for busy teachers: the week's most useful AI-in-education news, Australian stories first, each with a note on why it matters in your classroom. Free, and you can unsubscribe at any time.
