The government will create a £3 million “content store” to train artificial intelligence (AI) to be more reliable to help teachers mark work and plan lessons.
Government documents, such as curriculum guidance, lesson plans and anonymised student assessments will be pooled into a “content store”, with AI firms encouraged to use this to train their tools, the Department for Education said.
Ministers hope this will generate accurate, high-quality content, such as lesson plans and workbooks that can be reliably used in schools and colleges.
Stephen Morgan, the new minister for early education, claimed the announcement marked a “huge step forward for AI in the classroom”.
“This investment will allow us to safely harness the power of tech to make it work for our hard-working teachers, easing the pressures and workload burdens we know are facing the profession and freeing up time, allowing them to focus on face-to-face teaching.”
£1m to incentivise AI firms
The content store will be funded by £3 million from the Department for Science, Innovation and Technology (DSIT).
This includes a partnership with the Open University, which is sharing learning resources.
It is aimed at firms building tools to help teachers mark work, create teaching materials and to assist with routine administrative tasks.
To incentivise AI firms to use this, an extra £1 million will be awarded by the DfE to those with the best ideas to put the data into practice to reduce teacher workload.
Each winner will build an AI tool to help teachers with feedback and marking by March 2025, and applications open on September 9.
But none of this money, allocated as part of a wider DSIT project, will go directly to schools and colleges to help them develop and adopt AI.
DfE says providing AI with data boosts accuracy
It comes as the DfE said it would today publish test results showing that providing generative AI models with this kind of data can increase accuracy to 92 per cent.
This is up by a quarter, from 67 per cent, when no targeted data was provided to a large language model, it said.
The development of more sophisticated AI has sparked debate about its potential benefits to schools, along with fears about its potential misuse, such as if used to cheat assignments.
The DfE’s policy paper on generative AI in education also warns content created can be inaccurate, inappropriate, biased, out of date or unreliable.
In October 2023, the DfE hosted an AI “hackathon” event, where school and college leaders and tech experts discussed how the technology could be used to reduce teacher workload.
They explored how AI could help draft and review written policies published on school and college websites and how ChatGPT could be used to create parent newsletters, among other issues.
The DfE said teachers at its hackathons said standard AI tools “weren’t yet fit for purpose” for education use, as outputs were below national standards and they were tricky to use.
DfE to publish safety framework on AI products for education
The department also pledged today to publish a safety framework on AI products for education later this year.
Morgan will meet ed tech firms before setting out “clear expectations” for the safety of AI products for education.
In May, FE Week revealed ministers were planning to appoint ed tech evidence checkers to help schools work out which products deliver the best impact as part of an AI “training package” for teachers worth up to £5 million.
But this was put on ice when the election was called.
A YouGov poll of 1,012 teachers in the UK in November found almost two thirds think AI is too unreliable to assess students’ work or help with resource or lesson planning.
We’re told AI accuracy was 67% and they reckon it’ll be 92% with free access to education data.
A worthwhile question is what is the accuracy of humans doing this work?
(and the little imp in me would like to point out that if humans aren’t 100% accurate, then who / what calculated the 67% and 92%…).
My main concern is that this is a trojan horse – made to look like reducing teaching workloads, when it’s arguably really about reducing teaching workforce and saving money.