BY RICHARD TAYLOR
For someone with zero classical education, the ideas of the French philosopher Voltaire, such as Le mieux est l’ennemi du bien, can seem rather abstruse. I know a little about Winston Churchill, such as his quote, ‘Perfection is the enemy of progress’.
The ideas of both Voltaire and Churchill are relevant in the world of education as it grapples with the impact of artificial intelligence (AI). Quite how important the idea that perfection is the enemy of good/progress was brought home to me by the growing howls of outrage from lobbyists, influencers, parents and educators to the idea that AI has any place in education. A neat example of this furore can be seen in the binary Sky News story about David Game College which is using the ‘AI’ of CenturyTech with its GCSE students. Why this is even a story, given that CenturyTech have been claiming to use AI for over a decade (including that it can predict autism with 96% accuracy) is stark reminder of just how facile the coverage of education is in major media outlets.
This non-story of supposed innovation by a UK educational institution using AI then flips to the outrage of anti-AI (and tech in general) campaigners, such as Chris McGovern, Chair of the lobby group, Campaign for Real Education. Sky introduce McGovern as a former No.10 Education advisor, ‘who now campaigns against ideas like this’. His first words are, “children will lose a great deal from that AI experience ….the problem with the AI (sic) and the computer screen is that it is a machine…and straightaway you are dehumanising the process of learning”. This is an old argument that I have seen throughout my career in edtech and before, when we did school projects for Hewlett Packard (about calculators). It goes back as far as 1470 when the Florentine humanist Niccolò Perotti asserted that the books were so inaccurate it would have been better had they never been printed.
Those interested in education want to use technology to improve education in the unseen administrative work as well as assisting teachers to teach and students to learn. In the first instance (administration) there have been huge advances and benefits. Perhaps that’s why the Competition and Markets Authority (CMA) decided to pause their investigation this week into the activities of the owners of SIMS. For anyone outside education, SIMS is a system that many schools use for admin. At its height (when owned by Capita) it had over 80% of the UK market. It’s been argued that with its monopolistic position, SIMS held back the UK edtech sector for years, leading to the creation of competitors like Arbor, who were cheaper, nimble, cloud-based and user-friendly. Today SIMS has less than 44% of the market, still officially a monopoly (the CMA’s definition is 25% of a single market), but its continued decline may be why the CMA decided to park its investigation into what was alleged to be abuse of its dominant market position.
It’s in the domain of helping teachers and students that AI is most contentious and this is where the notion of perfect being the enemy of good/progress is so pernicious. Let’s look at a few facts:
- Teachers are expensive – median pay £43,801; with their 28.3% employer contribution pensions they are the largest cost in the £59.6bn education budget
- Overall, 85.9% of students are taught by a specialist teacher but this falls to 43.4% for key subjects like maths
- 43,522 teachers (excluding retirees) left the profession in 2023 with just 44,302 recruited. No surprise then that there is a chronic teacher recruitment and retention crisis. This won’t be solved by Labour’s claim that its imposition of VAT on private schools will raise £1.8bn to pay for 6,500 new teachers (which in theory will cost around £365m if you include the cost of pensions). To add to this tale of woe, 1 in 10 new teachers leave after one year and the target for secondary teacher recruitment was missed by 50%, something that occurred for 10 out of the last 11 years!
- The population bubble from the early 2000s is now hitting secondary schools. Numbers in primary are expected to fall by over 200,000 by 2028 but in secondary they are expected to rise by at least 29,000. There is some debate about whether primary numbers may fall by only 2% rather than 6% and the DfE’s top civil servant has writen to the Parliamentary Select Committee admitting the DfE has underestimated student numbers.
In practical terms this means there are too few appropriately qualified teachers for key subjects like maths and science and that things are unlikely to get better for some time, if at all. So with 56% of secondary maths students being taught by a non-expert, often inexperienced teacher, what do we do? Edtech, including that which uses AI, can help but it won’t fix the problem of getting more appropriately qualified maths teachers in schools. This is exactly where the problem of perfect being the enemy of good is so prevalent – let me give you two examples.
Speaking with the CEO of a successful tutoring company, I was shown a demonstration of their spoken one-to-one AI maths tutor, replicating the best practice and evidence of effective human tutoring. They developed it because recruiting highly-skilled tutors at any scale is as, if not more, challenging than teacher recruitment in schools. Given the size of the attainment gap and the reality of school budgets, human tutoring cannot solve this problem at the required scale. The CEO recalled a recent meeting with a senior non-UK educationalist who demanded that any tutors they hired had to be fully qualified and also registered teachers in their region. The CEO pointed out that if they could do this, which they doubted, all it would do is cause an even greater problem for school teacher recruitment! He also told me about a recent discussion with some top level civil servants (I won’t disclose where) who asked whether the company’s AI tutor was, “as good as a highly trained experienced maths teachers”. Of course it isn’t, was the reply, but that’s the problem of perfect being the enemy of good. It’s better that 56% of students in England who don’t have qualified experienced teachers could have access to pretty good, supplementary AI support.
I also spoke to multiple edtech founders, who described the recent £4m AI announcement by the DfE, who described the recent £4m AI announcement by the DfE, as as ‘confused policy at best’. I recommend that you read this document as it is equally optimistic about what AI might be able to achieve. For example, it specifically mentions a company called TeachMateAI (established June 2023), whose CEO claims, ‘TeachMateAI already saves teachers over 10+ hours of time each week through our AI tools’. Really? I have been involved in educational AI for several years and been very critical of the unevidenced claims about AI by companies like CenturyTech. Before this is put in official government correspondence, the DfE should have asked for robust evidence from TeachMateAI. When I asked them TeachMateAI’s response was that they have done a user survey with 770 respondents, which while encouraging, is ‘anecdata’ not robust evidence.
The debate about the merits and limitations of AI in education is a mess. The problem started years ago with unsubstantiated claims about what UK companies could do; it’s now shifted from scepticism and outright antagonism, fuelled in part but woeful media coverage (SchoolsWeek being a rare exception), made worse by an apparent intellectual, impartiality and policy vacuum at the DfE.
I don’t subscribe to the Prime Minister’s ambit claim that things will get worse before they get better, but in terms of education and AI, the fact that perfect is the enemy of good seems set to ensure this will be the case in England. It will hurt schools, families, students and UK plc, but I doubt anyone bleating about AI in education has given this a jot of thought.
Richard Taylor is a specialist in education sector start ups, edtech, business intelligence, market, company & sector analysis.