This is a transcript of the second ‘Meet The Education Researcher’ podcast that we’re putting out in August 2021 on the topic of AI and education. This episode features Dr. Val Mendes from UNESCO. Val brings a big picture perspective to the topic of AI and education, particularly in light of UNESCO’s focus on educational policy and practice around the world. In this episode, Val covered a lot of issues from what UNESCO is doing to support AI and education, through to the big questions of what AI can do well, and whether or not we actually need AI in the classroom at all. This transcript has been lightly edited for clarity.
First off, Val started by offering a basic working definition of what AI actually is. What’s useful here is his emphasis on seeing AI as technology that’s narrowly focused on certain tasks, rather than any sci-fi version of superhuman levels of general intelligence.
[VAL] In UNESCO – where I’m coming from – we have a definition that we worked up with an international commission. And this group described AI as machines capable of imitating certain functionalities of human intelligence. And I stress certain abilities of human intelligence. This can include perception, learning, reasoning, problem solving, language interaction, and even producing some creative work. Indeed, we have examples of music generated by AI – we had a big event recently here in Paris run by UNESCO where the music that was running was actually created by AI. But obviously, we also have the human part of it – and this is where we believe that the maximum use of AI is when we can have it as an element to support humans … and to have this augmented intelligence.
Following on from this idea of augmented intelligence, the next logical question to address is how AI applications are actually being used, and how AI might be used within different educational contexts. Here Val raises some interesting emerging applications of AI in education, and also makes a useful distinction between learning with AI, teaching with AI and system level uses of AI.
[VAL] I think we have many very exciting applications and many organizations are producing some promising practices – so we are trying to compile all these emerging experiences. But maybe it will be helpful if we classify at least three types of different applications of AI. So, the first is learning with AI. Here you have student-facing AI. The second, is teacher-facing AI. Here we have applications that are designed for teachers. Third, then, we have system-facing AI. Here are applications like education management information systems – that are built potentially into learning management systems – that basically are applications that integrate the learning processes of students and also have functionality for features. So if teachers can keep track of their students, get feedback, etc. And then we have big picture applications in terms of ‘data analytics’ that help decision making on the part of the government, school managers and directors.
So, this first layer of ‘learning with AI’ involves a lot of interesting tools. For example, we can detect if a student is performing very badly, and then we can quickly intervene – avoiding dropouts, etc. However, there are some downsides in terms of the ethical dimension – such as data protection, etc. The second layer of ‘learning about AI’ involves asking questions about how AI works, what AI is and what kind of skills do we need to know in terms of this technology? And here there are lots of very interesting open courses already on the internet. So this is about learning what AI is and developing understanding. Then, the last layer relates to preparing for AI – so, how we as a society or international community can be ready for the AI age? This involves putting in place frameworks and policy guidelines, because we cannot let this huge sector be a driven by specific groups or just the private sector. We are talking about more than 6 billion US dollars being generated by educational AI alone in the coming years until 2024. So, that means, there are lots of interests. I think here we need to get together and design together as humans in diverse ways the AI age that we want to live in.
Given Val’s current role, we’re also keen to get a sense of how UNESCO is developing its responses to AI and education. So what main lines of work is UNESCO prioritizing when it comes to AI and education? What are the emerging issues and problems that have been identified as requiring attention?
[VAL] Basically, I think from our perspective, what we are trying to do in terms of response sits at two levels. One is from the policy dimension involves working with governments, and conducting some research to see the actual impact. So, in terms of ethics, for example, we have been running some studies examining gender issues –for instance, why do Siri, Cortana, Alexa, and others have initially a female voice. Why are they always female assistants? So obviously, here are designs made by male, white designers that basically reinforce the stereotype of women as passive helpers, not leaders, etc. So this is one example. So we need to work with the industry – and we don’t work only with government – we bring all the main actors to the table, and we try to improve to improve their regulations, etc. This is one level. And then at the country level, obviously, we are concerned how these AI programs don’t work in very low resourced contexts. So here, what we need to do is work with the local ecosystems to see how schools, communities, local education software producers and local content producers can come together and design really contextualized solutions that can be – or not – empowered by AI.
Issues of gender, and the suitability of AI for low and middle income countries are clearly things that need close attention. So given these limitations, it’s interesting to hear from Val what types of AI he sees as having potential to benefit these context appropriate ways of using technology. In short, what does Val see as good examples of AI use with genuine potential to benefit education in low and middle income countries?
[VAL] I think the potential is huge, so let’s try to give some examples. There is one project that is called the Global Digital Library. This is a global initiative that is trying to provide access to school-age (K-12) books in different languages. So many children cannot access this content. So it was a human-created initiative, but now AI is coming to help the translation – so building chatbots and even conversational experiences into this tool. And this is a tool that we can bring to the classrooms, have interesting exercises with children, and somehow impact on learning. We also have a learning management system project with a private company and working with the Lebanese government to apply it for refugee camps. And we are using this tool to support learning and to track if there is a risk of dropouts, for example, and then suggest how we can intervene. However, I think we always need to be very critical as well – we need to remember to question what the purpose of using AI is for that specific pedagogic activity. It is important that the human teacher has the last words – he or she needs to design the learning activity in a way that connects with the context and with their students, and then decide which tool to bring. And for us, these tools really can empower learning – and that was the intention of our first year working on AI and technology. But I would say let’s add this critical view. Let’s also try to foster research to see the real impact on learning, and not just use AI for the sake of using it. And then we can bring more diverse actors – the way to avoid bias is to bring more diverse people to the creation of AI tools, and AI education tools as well. So we need to work from the very beginning with girls in STEM and STEAM – and to have girls able to code and to design the AI from today and from tomorrow. So with this in mind, I think we can make it work.
So a clear message here is that the technology on its own is not going to change things for the better, and in some ways, even runs a risk of exacerbating existing problems. Having focused on issues of gender and inclusion. It was also interesting to hear Val’s views on issues of pedagogy – another obvious issue that’s often rather glossed over when people talk about AI in education. Again, here, Val was keen to stress the need to bring diverse groups of people and ideas together to sort the problem out collaboratively
[VAL] We need to discuss the role of teachers, and also the pedagogy that AI is bringing to the table. So for example, if you look behind many of these systems that are automated let’s say an intelligent tutoring system – the kinds of underpinning pedagogy is a typical instructionalist pedagogy – a knowledge transmission style of pedagogy … the classic one. So I think if we really want to build intelligent AI systems in education, we need to bring other pedagogies that have been there for many years. You know, we are talking about collaborative learning … we do have some initiatives in AI that are doing this, but this is still not a reality. We need to bring guided discovery learning to the table, for example … the idea of ‘productive failure’. So you know, we need more complex pedagogies that can support learning in a specific moment. So, the thing is how do we do that? And to do this I think we need the teachers, we need the pedagogues, we need a very interdisciplinary and diverse group, building these educational AI tools … and then we might get it right.
This point about bringing diverse groups of people together to better design AI, also brings us neatly onto the question of how educators can work with the ‘big tech’ companies who are clearly at the forefront of developing AI. In particular, when we were discussing this, Val was keen to talk about the need to balance commercial interests against the idea of education as public good. Again, Val sees this as requiring us bringing together broad communities of industry, policymakers and public education.
[VAL] I think that some big tech companies and start-ups are working with international organizations as well as with communities. But I think the key issue here is to have the most diverse table as possible. So, we need to have these big companies, but also you need to have other members of the learning ecosystem. So together we can balance the interest and we can try really to contribute to the public good – to education as a public good. We understand that the private sector is a private business. But when it comes to education – and as UNESCO has been defending – education is a human right, education is a public good. So we all need to work towards this. So profits aside, we need to get this very clear … and then together we can improve the technology, because the challenges are huge. I think it’s important to bring the right critical view to balance the power that sometimes is skewed on the side of big tech and other companies. So, to do this, I think we need to work together on the relevant frameworks – we need to have legislation, we need to establish the rights of the student to control and own their own data. And how can we do this? We need regulation! So how can we improve the classroom and learning outcomes? Then we also can partner with whatever organizations we want to. But I think it’s a matter of balance and having all the actors together as I said earlier,
Practically achieving this balance is ambitious enough, but we concluded by talking a little more about other aspects of future potential for AI and education. So regardless of the policy and governance challenges ahead, what other aspects of AI in education is Val looking forward to coming to fruition in the near future,
[VAL] In terms of what AI can do well, for example, there was research by the OECD that found that only 11% of adults are above the level that AI is close to reproducing in literacy and numeracy. What this means is that AI is getting really, really smart in something that humans can do – that is, numeracy, reading and math, etc. So what is the implication? I think is it comes in terms of the skills that we are learning in schools – and here AI connects with the mission that we all have in education. We have more than 600 million students before COVID that were in schools but were not learning the basics of numeracy and literacy. So, how AI can help us there is a question that we now need to ask. And there are fantastic tools that we can bring to the classroom to try solve this problem. But I agree that there are other things that are fundamentally human – for example, how AI can actually feel a student’s emotions … we know of claims that the system can ‘read’ emotions … we can have all this face recognition etc. (even if that is not ethical, by the way). But a human teacher really can connect with this person … and engage in the communication skills, the social emotional skills that a machine cannot do. And I think these aspects of education should be with teachers – and not with AI.
So that was Dr. Val Mendes, talking us through some of the key issues currently surrounding the development of AI and education around the world. If you found this conversation interesting, then do be sure to check out the other episode of Meet the Education Researcher on the same topic, from the same panel discussion, featuring Professor Erica Southgate from the University of Newcastle. In the meantime, do check out Val’s Twitter feed – @ValMMendes