Elizabeth Holmes caught up with Miles Berry to discuss the benefits and ethical dilemmas of bringing AI into the classroom.
The potential role of artificial intelligence (AI) is something of a new frontier for schools. But should we not be finding out as much as we can, so that we can maximise the benefits and avoid the pitfalls?
Last December, the House of Lords Artifical Intelligence Select Committee heard from a group of three education experts, namely:
Together they gave varied evidence on current and possible future applications of AI in schools. The full session is well worth a listen.
What do we really need to know? Should we be fearful, cautious or excited? I caught up with Miles Berry to find out more.
Miles Berry: ‘The AI we have at the moment is still fairly rudimentary – most of the recent developments are in the territory of machine learning: broadly this is about software determining its own rules on the basis of the data it has, rather than just following the rules the programmer gave it initially.
'In the short to medium term, I think we can expect to see AI making teachers’ lives easier, letting them concentrate on the more human side of their role.
‘I think there are five broad areas where AI, in this sense, is already having some impact in education.’
Marking work
‘We're starting to see some applications of AI to assessment beyond the realm of multiple choice, numerical answers and short one- or two-word responses. EdX pioneered automatic essay grading, and we're now seeing this piloted for Australia's NAPLAN tests of persuasive writing for school children.’
Setting work
‘From the early days of AI we've seen rule-based adaptive learning systems, which tailor questions according to how well a student has performed on earlier questions. This approach is attracting interest at present in smart tutoring systems, which can use data from lots of other learners’ attempts at questions to suggest what might be good content or questions for any particular learner.’
Identifying possible interventions
‘Machine learning algorithms can quickly spot where data doesn't fit into the expected pattern, perhaps because a student is having an off day, or perhaps because of some special educational need. The system can then let teachers know that it might be worth having a chat with the student, or referring them for professional diagnosis if the special need had previously not been identified.’
Responding to queries
‘We've not seen much of this in school yet, but Georgia Tech pioneered the use of an AI teaching assistant, “Jill Watson”, to respond to students' questions – with very positive evaluations!’
Accessibility
‘We're already seeing AI applications make things much more accessible to students. Look at how:
‘I guess the benefit comes through the patience, and non-judgemental nature, of machine learning systems. AI can cope with presenting information and providing practice time and time again, when even the most sympathetic human teachers might find their patience running low.
‘Many students worry about making mistakes when answering a teacher's question or when handing in work for their teacher to mark: thanks, perhaps, to familiarity with computer games, students seem much less apprehensive about letting a computer mark their work or provide feedback on their spelling and grammar.’
‘I think this is pretty inevitable, unless as a society we choose not to allow it. If we think about how AI is already impacting on finance, entertainment and healthcare, I think we'll see widespread use of at least some aspects of machine learning in education quite soon – initially as tools to support teachers in the more easily automatable aspects of their role – for example setting and marking work.
'We'll see some use of AI for individualised tutoring, perhaps initially for some special educational needs and for niche subjects where the school doesn't have any qualified staff.’
'Eventually is a long time, but I think there's plenty that teachers do that the machines won't be able to do at all well in the foreseeable future. On this list are things like:
'There are some great introductory books available. I'd recommend Luke Dormehl's Thinking Machines (published by Tarcherperigee) and Pedro Domingos' The Master Algorithm (published by Allen Lane) as good starting points.
'One readily accessible approach to experimenting with AI in school is to set up a Google Assistant, Apple Siri or Amazon Alexa in class and start getting it to answer some of the questions that come up in a lesson.
'Going a step further, Raspberry Pi's AIY kit lets teachers and students create their own new skills for Google Assistant.
‘For those teaching computing, whether in primary or secondary schools, the best way to get to grips with AI is to have a go at teaching it: the national curriculum for computing provides ample opportunity to do so, although there aren't yet the materials we need to support teachers with this yet. One way in is through building (or using) simple machine learning classifiers, another is through experimenting with creating chat bots.’
One accessible way to experiment with AI would be getting virtual assistants to answer questions in a lesson
‘At the moment, the field is largely unregulated – just as the Warnock Committee laid down the ethical framework for IVF and embryology, we urgently need a similar body to set out the ethical principles for AI. The European Parliament has done good work here, but a proper ethical and legal framework is essential.
‘I think those developing AI systems need to be honest about the assumptions they make and the decisions they take. One of the challenges with machine learning as it stands is that many current approaches make it impossible to understand the rules the machine has developed for itself on the basis of the data.
'Imagine the parent consultation:
Teacher: “The AI says your daughter should do this course.”
Parent: “Why?“
Teacher: “Nobody knows.”
'We're used to being able to explain our decisions, and somehow "the algorithm says so" doesn't really cut it.’
‘Longer term, there are big ethical issues facing us as the machines take over more and more human activities. When computers give every appearance of being conscious, creative entities, what rights and responsibilities would they have? How will we and they co-exist?
‘In a utopian, post-scarcity society, if, in a few generations time, there's nothing left for humans to do, what then should we teach in schools, or perhaps what should the machines teach in schools? If education was no longer, even in part, about getting a job, how would that change school? What would students study if there was no longer a need to study?’
‘England's national curriculum has placed a renewed emphasis on knowledge and traditional subject disciplines. Its creators have taken the view that the national curriculum needs to "set out only the essential knowledge that all children should acquire". Its stated aim is to provide pupils with an introduction to the essential knowledge that they need to be educated citizens. It introduces pupils to the best that has been thought and said; and helps engender an appreciation of human creativity and achievement.
‘At the root of this is the notion that a person's ability to take part in and contribute to society is dependent on how much they know – participating in debate, making choices and even reading a quality newspaper are much easier if you've a good level of all-round general knowledge.
‘How this vision is realised is, to a greater or lesser extent, left to schools and teachers to determine for themselves. Many schools and teachers see their responsibility as extending beyond the knowledge core of the curriculum with the opportunities to develop the character, curiosity and creativity that are likely to matter at least as much as the facts when it comes to adult life.
How this vision is realised is, to a greater or lesser extent, left to schools and teachers to determine for themselves
‘Others have taken the message of the new curriculum to heart and are now prioritising knowledge in classroom practice, with lots of quick question and response to help with automatic recall, knowledge organisers to ensure pupils know what they need to know, and homework designed to reinforce what they learn that day, or in preceding weeks.’
Miles Berry is principal lecturer in computing education at Roehampton University. You can follow him on Twitter @mberry