New artificial intelligence (AI) technology could solve teaching’s recruitment crisis, end formal testing and ease teacher workload, leading scientists claim.
But they also warn that unless teachers are educated about the technology, it could fall victim to false marketing and pale imitations.
Rose Luckin, professor of learner-centred design at the UCL Institute of Education, said that technology already exists that not only identifies students’ academic strengths and weaknesses, but also their levels of confidence in their learning and what motivates them. This information could be used to provide one-to-one tuition that responds effectively to individual pupils’ needs.
“I do not believe that it’s about the technology replacing the teacher, but I do believe that roles will change,” Professor Luckin (pictured, inset) said. “Think about shortage subjects. It’s about saying, ‘It’s OK. I’m not an expert in science, but I have a system that is.’ You can use the computer technology for what it’s good at: being able to cover subjects in depth, even when a particular teacher doesn’t have that subject expertise.”
As AI technology allows for personalised tuition, it can also provide CPD for teachers.
“If you have someone who has maths A level, but isn’t a qualified maths teacher, they can have a computerised companion who helps them to develop their skills,” Professor Luckin said.
“It would also help with teacher burnout, by providing assistance and support.”
Sir Anthony Seldon, vice-chancellor of the University of Buckingham and former master of Wellington College, is writing a book about AI, entitled Fourth Education Revolution.
“AI will certainly help with schools that are struggling to find good teachers,” he said. “The level of maths or physics or history knowledge will be high. Also, they will be able to adapt to students’ learning needs in a very individuated way.”
Computerised adaptive testing already exists, adjusting test questions according to pupils’ responses to previous questions.
But Professor Luckin says that AI technology is vastly more sophisticated, responding not only to pupils’ previous answers, but also to their levels of confidence and motivation.
She suggests that the data generated by these programmes could eliminate the need for formal testing.
“Once you’ve got that kind of detail about the pupils’ progress, why do you need to make any kind of summative assessment?” Professor Luckin said. “You can show how tenacious someone is, how much help they need.
“If you’ve got this kind of detailed data about the individual learner, then you can extrapolate from that how the whole class is performing, and the whole school.”
The US military, which has funded much of the research into AI in education, already uses the technology to train navy IT specialists to solve complex problems.
But classroom AI technology is now being developed by US companies such as Carnegie Learning and Alelo. Professor Luckin estimates that, within the next three years, it will be available to all schools.
Ken Koedinger, a professor of human-computer interaction at Carnegie Mellon University in the US, says that some schools in the country already use AI technology to help with one-to-one learning. It functions in the same way that school language labs did: pupils work at their own pace, but can call on the teacher for help when they need to.
Teaching unions are sceptical about the idea of AI solving the recruitment crisis. Russell Hobby, general secretary of the NAHT headteachers’ union, pointed out that the technology had come too late to help with today’s problems.
“It’s a supplement to the teacher,” he added. “It could help them do their job better, but it won’t replace them.”
Mary Bousted, Association of Teachers and Lecturers general secretary, dismissed the idea that AI might prove to be the solution to teacher shortages as “a bit ambitious”.
She said: “What we’ve got to do is think about how we make teachers lives’ better, so that more people want to go into teaching and to stay in the profession. There’s no getting around that.”
Professor Luckin and academics from education company Pearson have recently published Intelligence Unleashed, a report on AI in the classroom (bit.ly/IntelligenceUnleashed). She insists that it is vital for teachers to understand enough to know what they are spending their money on. Otherwise, she fears that AI could become the next learning styles: scientific nonsense marketed to teachers as truth.
“There’s a lot of stuff that – I wouldn’t say ‘snake oil’, but it’s claiming to use AI when it doesn’t,” Professor Luckin said. “That’s what really worries me: that money will be spent on the wrong thing. It will be learning styles all over again.”
Turn to page 26 to read more about robots in education, in this week’s TES feature
Will AI replace teachers?
Could AI put teachers out of their jobs? All education researchers in this field are unanimous: absolutely not, they say.
“It’s more a case of teachers and learners having a whole host of different resources available to them. Some human, some machine,” said Rose Luckin, of the UCL Institute of Education.
“What we want is to complement the human teacher. Because you’re not going to build a system that can do everything a human teacher can. But I think we have to accept, as humans, that our roles are changing.”
Sir Anthony Seldon believes that AI will revolutionise society as effectively as the printing press once did. But, he added: “An algorithm can never replicate human feeling. So that’s why we need to have the teacher.
“AI will allow individuated intellectual learning, and will give a proximate emotional experience. But it won’t be a real emotional experience. It will be a simulation of a human relationship, but not a human relationship.”
How do teachers differentiate between genuine AI technology and pale imitations?
There are two sets of questions to ask when deciding if AI technology is genuine, Professor Luckin explains.
Firstly, the answer should be yes to the following. Does the technology:
Make fine-grained adjustments to support learners – ie, adjustments that involve micro-steps required to complete a task, not just right or wrong answers?
Maintain an updatable, detailed model of the subject knowledge that the learner is studying?
Keep a detailed learner model that’s updated after each interaction and influences learning activities and support provided by software?
Retain an updatable and detailed model of the teaching and learning approach adopted to support the learner, which influences learning activities and support provided by software?
Secondly, ask for evidence that the technology has helped learners improve their knowledge, skills or attitude.
The product or software should have been evaluated with learners who are similar to your students, in a setting that is not lab-based. Data collected in this evaluation should be analysed to show not just that learners’ knowledge, skills or attitudes have improved, but when improvements have been recorded, which features of the technology have been used and how.
The evaluation should preferably have been conducted by an independent organisation.
What can AI do?
There are three main components of AI education technology:
Knowledge of the subject being studied.
Knowledge of individual learners, built up over time to create an understanding of how they respond to tasks, how flexible they are, whether they have stamina and whether they hesitate over certain tasks.
Knowledge of the processes of teaching and learning.