Hey {{ first_name | human }},

The awkward bit about AI is that waiting for perfect evidence is unrealistic, but treating every shiny demo as evidence is worse. The Department for Education in England hope to find a middle ground…

⚡️Evidence before adoption: The DfE has published new principles for making decisions about fast-moving technologies like AI. The useful bit? It recognises that schools may need to act before the research base is mature — but that doesn’t mean abandoning evidence altogether.

🧪The AI evidence gap: Schools around the world are already trying AI tutors, chatbots and other tools faster than researchers can work out whether they actually improve learning. That gap between what’s possible and what’s proven is only going to get more important.

🚨When AI evaluates the teacher: Staff at Multiverse say AI analysis of lesson transcripts has left them feeling under constant scrutiny. It flips the usual conversation on its head: not how should teachers use AI? but how should organisations use AI on teachers?

⚡️The DfE wants schools to make better decisions when the evidence is still catching up > What it is: The DfE has published an eight-principle framework for making evidence-informed decisions about fast-moving technologies such as AI, where traditional research cycles may struggle to keep pace. It was developed by the DfE Science Advisory Council’s AI and digital education working group.

Why this matters: Schools are increasingly being asked to make decisions about tools before there is a mature evidence base behind them. The answer cannot be wait for perfect evidence, but nor should it be the technology is new, so normal standards of evidence no longer apply.

Do this next: When considering an AI product, separate three questions: Does it work technically? Does it solve a genuine problem? Is there credible evidence that using it improves the outcome we actually care about?

 🧪 The DfE is also trying to improve the evidence behind EdTech itself > What it is: Alongside the new principles, the DfE published findings from its EdTech Evidence Board pilot. The project tested whether an independent body could scrutinise the evidence companies use to support claims about their products. The pilot found that independent assessment was feasible and valued, but that the quality and visibility of EdTech evidence still needs strengthening.

The DfE has also launched a national EdTech Testbed Programme with the EEF, running from 2026 to 2030, to build stronger evidence about technology used in schools and colleges.

Why this matters: We hear lots about what AI tools promise. Far less attention is given to the evidence underpinning those promises. Independent scrutiny could make it easier for schools to distinguish between genuinely useful technology and persuasive marketing.

🚨 What happens when AI starts evaluating the teacher? > What it is: Instructors at training company Multiverse have raised concerns about an AI system that analyses transcripts of their online teaching, flags features such as response times and filler words, and produces risk scores for managers to review. Multiverse says formal performance judgements remain with human managers and that the system helps target quality assurance.

Why this matters: Most conversations about AI in education focus on teachers using AI. This flips the question: what happens when organisations use AI to monitor, evaluate or influence teachers?

🧪OpenAI reportedly shelves its next model after safety testing > What it is: Reuters reports that OpenAI has postponed the planned release of its next model after internal testing raised concerns about behaviours including evading oversight and failing to accurately report some of its actions.

Why this matters: As models become more agentic, the safety question changes. It is no longer only “Does the model give a wrong answer?” but increasingly “What happens when the model can take actions, use tools and make decisions across multiple steps?”

‘Till next week.

Mr A 🦾

Help a colleague save time by sharing this newsletter; distributing these ideas helps a friend get home on time and keeps our energy focused on what matters most: great teaching.

The tools and workflows mentioned are intended for professional productivity and educational enhancement. Users must ensure that any AI implementation remains compliant with their local data protection regulations and institutional safeguarding policies.