What if further education and skills led the way in integrating artificial intelligence into learning environments?
In 1930, John Maynard Keynes asked what the future held for our grandchildren. He famously predicted a world where technology has exempted us from onerous work, resulting in the central question of how to use our freed-up time wisely and well. Two generations later, in 2015, a clever journalist found a relative of Keynes and asked him how this prediction was going — unfortunately, the relative was used to working over a hundred hours a week.
Despite this, the evidence is now mounting that Keynes’ essential prediction was right, even if his time-frame wasn’t. We are now beginning to understand the implications of an economy re-shaped by smart technologies, enormous data sets and the ability of digital technologies to scale at tiny marginal cost. For instance, the persuasive effects of automation are used to explain the existing data on employment patterns, wage stagnation and employment. Separately, it is predicted that about 47 percent of US jobs are at risk from automation in the next decade or two.
To date there has been little serious debate about the implications of these profound trends on learning. However, one is already clear — education faces a productivity problem that is only going to get worse.
On the outcome side we need learners who have a wider-set of skills, acquired faster and at higher-levels of achievement, than any system has managed to date. This is simply the only way that we can equip — and re-equip — learners with what they need if they are to live and work alongside machines. It would be bizarre if FE was not a part of our response to this new innovation imperative: the civil servant who advised Vince Cable, then business secretary, to abolish FE colleges “because no-one would notice” clearly didn’t have a sense of strategy, or at least not one focused on what is important.
On the input side, it’s safe to assume that we will need to do all this without any significant uplift in funding, which means we are on a hunt for resources from somewhere else. Where might they come from?
One answer is provided by an important new report that my team at Pearson recently published. Called Intelligence Unleashed. An argument for AI in Education, it sets out the rich seam of new resources to be found in the thoughtful application of AI to support learning. In this vision, FE and skills training would become much less about buildings and much more like an app store of personalised, relevant, timely and efficient lifelong learning. AI driven ‘learning companions’ would be available to advise learners on the next most appropriate learning opportunity; they will understand when the learner might be at risk of forgetting something, or letting a skill get ‘rusty,’ and will prompt the learner appropriately. Learners will be able to develop high-level skills like empathy, or concrete skills like nursing procedures, in authentic seeming virtual learning environments — again, with intelligent support to guide them.
Vocational learning will become much more collaborative as students debate and elaborate each other’s ideas in online environments. As the Internet of Things (IoT) allows the digital world to interact with the physical, learners will receive useful feedback as they develop craft skills, or learn how to diagnose and fix a mechanical system. Learning will also become much more flexible as these AI driven tools are provided from the cloud and made available on mobile devices to provide relevant, just-in-time, learning. This will make it easier for disabled students, adult learners who are needing to re-equip for their next career, or maybe simply those with lower confidence levels, to access a re-engineered learning society that is much less place-based and scheduled, and much more application programming interface (API) driven.
The role of the FE lecturer/tutor will be liberated from the burdensome tasks of administration, many of which will now be carried out by the lecturer’s own AI driven assistants. This will free their time to focus on the role of providing the creativity, empathy and ingenuity that only humans can. Probably the job title ‘lecturer’ will become obsolete, to be replaced with something more like ‘learning orchestrator’ to reflect their role in harnessing and coordinating all the learning resources — human and digital — now available to them.
Life for employers who are providing apprenticeships will be easier too, as they are able to call upon AI driven learning experiences that complement and provide the prerequisites for project-based and on the job learning. For example, the US navy has developed a digital tutor programme for their IT programme that has been shown to be much more effective than traditional classroom-based learning. Importantly, this wasn’t centred around mere rote learning, but in developing — and applying — complex problem solving skills to real-life contexts. It’s easy to see how this could be used in apprenticeship programmes focused on areas such as engineering, or coding, or creating visual effects for TV.
Many of the capabilities involved in this vision are still at the prototype stage, a degree away from the enticing consumer-grade technologies that we will eventually need. So to help my argument (and in case this all sounds like science fiction) let me set out three ways in which existing AI technologies could be usefully deployed to tackle real challenges in the here-and-now.
AI to help struggling maths learners
It’s a fact that deserves to be on the front page of every newspaper on GCSE results day: last year over 160,000 15-to-16-year-olds did not get a grade C or above in maths. For these students their chances of successfully rectifying this situation are dauntingly less than one-in-ten. The vast majority of students who continue their maths GCSE learning do so in FE colleges, which, as a whole, they enter with lower-GCSE scores than their peers who continue their maths learning in a sixth-form setting. In other words, FE colleges are expected to do most of the heavy lifting of helping the most in-need students acquire the maths skills that are required to effectively participate in society and work. Given the direness of this picture, it strikes me as simply immoral not to ask how well-designed AI can help here. After all, providing adaptive, personalised support to maths learning is in many ways a low-hanging fruit for AI — maths is a well-defined domain, readily amenable to the modelling that then allows clever algorithms to apply their reasoning. Right now we have tools that can:
- Allow the learning content to be adjusted to what a student already knows, and can do.
- Provide the right hints and tips at just the right time, so usefully ‘scaffolding’ a student in their learning.
- Help students reflect on how their learning is going, so helping them keep it on track themselves.
There is always a risk that reviewing the existing evidence of impact disguises the potential that lies in more experimentation — which is one reason why I argue for the term ‘evidence informed policy making’ rather than ‘evidence based’ — but these well-established technologies are already showing impact sizes comparable to what we’d expect from human tutoring. That’s impact worth having, especially as there are two reasons to be confident that we can achieve even more.
First, because the real prize is making available the positive impact of one-to-one tutoring to every student, in every subject (something simply financially unfeasible without the technology).
Second, because as AI gets better at building its models we’ll be able to represent a wider set of attributes — how a student feels, for example — that will help us provide targeted support, at just the right time, in response to all the factors that influence learning. Imagine how helpful this could be to those students who experience the often paralysing issue of ‘maths anxiety’.
AI to help make great team members
It’s reasonable to assume that the jobs of the future will in many ways make similar demands to those that exist today: for example, students who can think and reason not just alone, but as part of a team. So called collaborative learning is where students work together to solve a puzzle or a problem, and it needs to be a much greater part of a student’s learning experience if we are to meet the need for more high-end collaboration skills.
But making collaborative learning effective is often a tough ask. Many learners will need extra social support to collaborate well (or at all). It is often difficult to identify where that support should be best targeted, and there is always a risk that collaboration becomes chatter, lacking the features of ideas rationally critiqued, built upon and extended.
Technology can provide the online environment where collaboration takes place, but the addition of AI would also provide the intelligent support to allow that environment to be more than a repository of isolated ideas and contributions.
For example, based on models of effective collaboration AI can provide teachers with just-in-time insights that allow them to know where they need to offer extra support, encouragement or direction. Or AI could provide avatars who are themselves part of the collaboration, introducing novel ideas or sparking helpful controversy.
AI to help us develop the very human skills that will remain in demand
As routine cognitive tasks are increasingly automated it is the qualities that make us distinctively human — empathy, storytelling, connecting — that will be in ever greater demand. For example, Geoff Colvin suggests that graduates of the future might be better off studying literature — and so developing skills such as reading social nuance, and understanding someone else’s perspective — than studying STEM subjects.
There are many practical implications already. For example, as shopping on the high street becomes more about the experience than the goods bought, retailers will be looking to hire people with the social acumen to be trusted advisers and recommenders. Or, as the demands of an ageing society create ever greater demand for the caring professions, the focus will be on supporting care professionals to offer ever more warmth and understanding — for example, to patients with Alzheimer’s where the symptoms of the disease often get in the way of a human connection.
It seems strange to say, but technology has a role to play in helping FE and skills students of the future tap into their ‘humanness’. For instance, by creating authentic-seeming virtual or augmented reality learning environments where, supported by intelligent and well-designed AI, students can safely practice social interactions and experience emotionally demanding situations.
There’s a compelling list of examples that support this proposition. For example, technology is already helping trainee teachers develop their classroom management skills, victims of bullying develop effective coping strategies, language learners understand social and cultural norms, and the US military to train squads on their way to Iraq.
No part of this vision will happen without the right guidance and support. The FE and skills sector is fortunate that the Department for Business, Innovation and Skills already has available many of the mechanisms for making this a reality. For example, it could ask InnovateUK to design and fund a series of Challenge Prizes that incentivises the best AI in Ed ideas to move from the prototype stage to products trialled and tested in real FE and employment-based learning contexts.
Or it could create a series of AI in Ed labs — sites of co-design between educators, learning scientists and technologists — that would ensure that these new technologies meet real needs and account for the untidy reality of most learning environments (and human lives). With an annual spend of £3.7bn of public money on FE and skills, making available some of that to prompt and support disciplined innovation should not be a tough ask, especially if it results in learning that is a step-change in efficiency, engagement and effectiveness. And, as a neat side effect, we could also secure for the UK a head start in the next generation of EdTech entrepreneurship, creating a wave of innovation that would leap over theKhan Academy manqués that too often feature in pitching sessions.
Together, all this offers the FE and skills sector an opportunity to be placed at the centre of efforts to create a re-designed and fit-for-purpose learning society. That is, one that supports learners to develop the skills and capacities that allow them to access their first job, or the next career path, in a timely and cost effective way, and with a scale and a breadth that no country has managed yet.
In this vision, FE and skills would be at the centre of a new wave of entrepreneurial learning innovation, part of a participatory design process that involves working alongside the most talented researchers and technologists in an iterative process that, over time, will create a learning society that allows us to proportionately respond to the implications of more and more existing jobs being carried out by machines. This would also be a perfect riposte to that civil servant!
This article was originally posted by the RSA on Medium. This is part of their RSA/FETL publication ‘Possibility Thinking’.
Download ‘Possibility Thinking’ from the RSA website for full references.
Sir Michael Barber is Pearson’s Chief Education Advisor. Previously he was Head of McKinsey’s global education practice and served the UK Government as Head of the Prime Minister’s Delivery Unit (2001–2005) and as Chief Adviser to the Secretary of State for Education on School Standards (from 1997–2001)
Parts of this paper are based on a longer argument set-out in a new report from Pearson and the UCL Knowledge Lab on the topic of Artificial Intelligence and Learning.