Prashanth Shenoy researches both halves of that problem: what teacher AI training actually changes, and what the EU AI Act now demands of schools. Published evidence, two free tools, and findings put plainly enough to act on.
teachers across 19 countries, synthesised in one meta-analysis of AI adoption
teacher responses from hands-on AI workshops run in Bengaluru classrooms
reviews of educator AI-literacy frameworks that address explainability at all
where the EU AI Act puts admissions, assessment, placement and exam monitoring
Every answer comes from a published study below. Predicting before you are told is also how people actually learn things, which is rather the point.
A 90-minute hands-on AI workshop for teachers. Of five measures of readiness, how many actually shifted?
Only ease of use and self-efficacy shifted. Usefulness, enjoyment and intention were already near the top of the scale before anyone walked in. The room did not need persuading; it needed fluency.
Across 28 reviews of AI-literacy frameworks for educators, how many explicitly address explainability?
Explainability is what decides whether a teacher can trust an AI judgment about a student. It is almost entirely absent from the frameworks schools are currently adopting.
India's school AI policy, scored 1 to 5 against UNESCO and OECD benchmarks. What did data protection score?
The curriculum scores full marks. Data protection has no education-specific provisions and no standards for the AI tools already running in classrooms. Curriculum-forward, governance-light.
38 studies, 15,919 teachers. How well does the standard adoption model predict what will happen in your school?
The averages are strong and well estimated. But the range within which a new school could plausibly fall runs from below zero to almost one. Useful as a compass. Useless as a map.
The finding first. Open any of them for the full explainer, the charts and the citation.
Two further field studies on AI literacy in Global South classrooms are working through peer review. If the current numbers would be useful to you, ask and I will send them.
Both are free, both work today, and neither needs anyone to get in touch first.
Find out how much training your teachers actually need, before you book any.
Most cohorts are already convinced AI is useful. Knowing that in advance changes what a session should spend its time on.
Something new about AI, every day, in a few minutes.
The research reason it works this way: one workshop moves usability confidence and very little else. AI does not sit still long enough for a single session to hold.
Keynotes, teacher AI literacy training, leadership briefings, full-day workshops, and expert comment for media. Tell me what you are trying to solve and I will tell you honestly whether I am the right person for it.
Prashanth Shenoy is a Grant Researcher in the HUMBLE HCI group at the University of Jyväskylä, Finland. Most people he works with call him PS.
The work runs on three tracks. Field studies take generative AI into Indian classrooms and measure carefully what a short intervention does and does not change. Synthesis work pools the international evidence on teacher adoption and tests whether the field's favourite model still holds. Governance work reads the EU AI Act against how these systems actually behave when a student is sitting in front of one.
All of it is co-authored with Mirka Saarela at the University of Jyväskylä, and supported by the Research Council of Finland.