Rose Luckin, Joshua Underwood, Kaska Porayska-Pomsta, Lewis Johnson and Lee Ellen Friedland
From HAL in Space Odyssey, through C-3P0 and R2-D2 in Star Wars, to Sonny in I Robot, Hollywood has been good at making money out of our fascination with machines that think and behave intelligently. And it’s no longer science fantasy.
We have computers that can fly planes, model countries’ economies, search the internet and predict what we want to type into a text message. We’re also developing computers with human qualities such as the ability to understand language and recognise visual images.
Education can take advantage of all this progress. Tailor-made learning is within our grasp as artificial intelligence (AI) empowers computers to deal with the fact that everyone is different. We differ physically, emotionally and cognitively – and in our ability to understand how we learn and when we need help.
An education that recognises these differences can help everyone achieve their potential. Such personalised learning requires teachers, tutors, parents and mentors to ensure that every student works on problems that are appropriate for them, problems that stretch them and help them progress. Software that uses AI can help ensure that learners receive relevant feedback, whether working individually or as part of a team. It can give them valuable information about their performance, enabling them to manage their own learning and emotions.
Education systems with AI are very adaptable. They can respond quickly and appropriately to information about what the aim of a lesson is, who the students are, who is working with whom and where it is all happening. And they can do this even if the information changes or is incomplete. The capacity to adapt to students’ abilities, needs, circumstances – even their moods – is underpinned by sophisticated AI techniques.
There are three main ways in which AI techniques are used to develop ‘adaptive software’ systems:
Building computer models that can act as scientific tools
ThinkerTools is a microworld that allows 10 to 14-year-olds to test their ideas and understanding of forces and motion. Students can run simulations of objects moving and observe how various forces such as impulses, gravity, and friction impact on these objects. The software can be set up to run according to Newtonian laws, and also according to other laws of physics. Students can run existing simulations or create entirely new microworlds, including game-like simulations with targets and timers.
Enabling a learning environment to adapt to input from learners, teachers or others
Andes Physics Tutor is an intelligent homework helper, popular in the United States. Students are presented with a physics challenge, requiring them to, for example, draw vectors or coordinate systems, define variables or write equations. Andes provides them with feedback every step of the way and encourages them to use helpful problem-solving strategies. It also changes its advice in response to the kind of error the student makes.
Designing computer models based upon a particular theory of learning
The Ecolab software simulation environment is intended to help eight to 10-year-olds explore food chains and webs. It is based on our understanding, courtesy of Russian psychologist Lev Vygotsky, that children make good progress when their learning is ‘scaffolded’ or supported by a skilled adult.
Imagine learning how to argue about the standard of your accommodation in France by standing in a hotel lobby having a heated conversation with the manager. Not only would you acquire the words, but you would also assimilate the body language and pick up tips on how to behave in such a situation. That, in essence, is the kind of lesson offered by Alelo’s ‘virtual-world simulations of real-life social communication’.
Alelo’s Operational Language and Culture Training System uses a virtual game based environment and interactive lessons to provide foreign language and culture training. AI techniques create convincing social simulations that can process students’ speech and behaviour, engage in dialogue and non-verbal interaction, and evaluate their performance. Independent evaluations have shown significant gains in students’ knowledge of language and culture and greater self confidence in their ability to communicate.
Andy is adept at using his intelligence to improve the social skills of young children, particularly those on the autistic spectrum. Through interacting with this virtual boy who inhabits a touch-screen magical garden, five to seven year-olds are encouraged to practise skills related to ‘joint attention’. This crucial skill, by which one person makes another aware of an object or event by pointing or looking at it, is often missing in children with autism.
Andy plays with the children, coaxing them to, for example, pick flowers or stack pots. Thanks to AI modelling, he can ‘see’ his young users, reason about their actions and, crucially, tailor his responses to them in the light of his observations and inferences.
Andy is the invention of Echoes, one of the Technology Enhanced Learning research programme’s projects18. Echoes researchers have equipped him with an underlying personality that influences his actions – much like a human, albeit in a simplified way. This means he can emulate, using AI techniques such as planning, at least some human behaviours such as having goals and acting on those goals based on his understanding of the current state of the world.
Andy’s built-in AI allows him to have credible interactions with children – and they generally enjoy the opportunities he gives them for exploration and experimentation. They also enjoy having a go at the challenges he sets, particularly the immediate feedback minus any real-world consequences. The fact that the scenarios can be repeated endlessly gives them both pleasure and a sense of control over the Echoes environment.