"I'm very bullish on the use of AI as a partner in the instructional process"

Vijay M.S. Kumar is a leading expert on education and technology. He was director of the Office of Innovation and Technology and associate dean for Open Learning at the Massachusetts Institute of Technology (MIT). He has spent 27 years building his career there and has become a key figure in the field of educational innovation. His experience makes him an ideal choice to lead distance education initiatives. In this role, he helped the Universitat Oberta de Catalunya (UOC) take its earliest steps as the first 100% online university and has been an advisor ever since.

During Digital Universities Europe organized by Times Higher Education and the UOC, he and Teresa Guasch, the UOC's Vice Rector for Teaching and Learning, spoke about educational innovation and technology, the evolution of teaching methods and the challenges posed by the new tools emerging.

As you have noted on occasion, what you like about MIT is the combination of looking both over the horizon and under the hood, and the fact that it tends to focus on the toughest problems. How do we effectively organize and manage educational innovation to make it effective and attractive yet also manageable?

If students are not learning a particular thing the way we want them to, do they have conceptual misunderstandings that we can correct? Should we change the learning experience? All of these are subject to innovative possibilities, which is essentially saying can we change the input parameters in creative, imaginative ways so that we can get better outcomes? There's a lot of things you can do for that. You shift chairs in a classroom and have a different element, that's an innovation. Can I use a little bit of technology to create some applets for visualization? That's an innovation. Can I use AI for better learning opportunities? That's also an innovation. But the underpinning over there is that you look to say, how can I imaginatively use the resources that I have to arrive at different learning outcomes?

We are at a time when we have such immense possibilities. If you look at a course or what we teach, every aspect of it is prone to innovative possibilities because of digital technologies. A lecture is no longer a lecture. It's a video, which can be immediately followed by some formative testing, formative assessment – if you don't understand the concept, suddenly you can have a visualization. So you have changed all the parts of delivering a lecture into these different kinds of components. And you're doing all this because you're learning something from learning science about how to do it better. Part of innovating is that I have to experiment with it and assist it. So I have to experiment and honestly assist and make corrections, not just in how they [the students] learn alone, but how I teach also.

The other part of it, you said innovation, and it's a complex term because often we confuse the invention with the innovation. For me, my research work is an innovation diffusion. To apply an innovative intervention and invention, I need to think about what does the classroom look like, what does the infrastructure look like, is the teacher prepared to do that, are the learners prepared. All that is in the realm of the innovation. It's not just about you, the teacher and the learners, it's also about the ecosystem. At MIT, we place a lot of value on active learning, hands-on learning. Hands-on practice, they can actually do through an internship in an industry. I care about it, but maybe I'm not the best person to deliver that. So you can look at innovative combinations on the supply side to deliver the outcomes also. [It's about] inquiring about what's going on, experimenting, assessing. And then thinking that you have to really look at it through a systems approach. That's the orientation that makes us very innovation-pro.

That's great because it allows us to assess the entire process. And your questions are good ones, because MIT is all about innovation.

Many years ago, when we launched OpenCourseWare (OCW), in the US, there were some initiatives that were being launched to commercialize educational content. [People at MIT were saying:] "What are we doing about the internet? Have you missed the opportunity?" And our faculty said: "we have to really think about what is the quality of excellence at MIT". At that time, there was a lot of discussion. Words like intensity were thrown around. And people said the real value of quality of excellence at MIT is a very high bandwidth of interaction between great faculty and great students, because we take a lot of trouble recruiting great faculty, we take a lot of energy getting great students.

And we are a research institution, and we want to bring the practice, the tools, the joy of research into the teaching-learning experience. This notion of experimentation is very much in the research DNA of MIT. Let's try it, let's measure it, let's assess it. And then if it's not good enough, try to correct it or throw it away. So, when we launched OCW, we clearly said this is a publication of MIT courses, it is not an MIT education. Because all that stuff, interaction, research, we know only how to do in face to face.

You mentioned artificial intelligence, but the emergence and widespread adoption of generative AI in higher education entails seriously reconsidering the processes involved in teaching and assessing students' acquisition of knowledge and competencies. At the UOC, we think of it as an opportunity, but how do you think universities can adapt to this new context? How should we rethink teaching and assessment, and how should assessment models evolve in this context?

My PhD is in Future Studies in Education, and I got introduced to some wonderful work that was happening in AI those days. There was wonderful work going on at MIT. I was at UMass Amherst and I had to come to MIT to see some of the work that was going on. We were designing intelligent tutors in the early days, and there were programs, there was one called Buggy. Buggy was to see how kids were doing a math program, and to see the pattern of mistakes they were making, because the patterns of mistakes reflected the conceptual misunderstandings that they had. It's a very interesting use of AI. AI helps surface those kinds of things, so that you can intervene appropriately. You can say, maybe I need to change my teaching strategy, or get to some other kinds of experiences that will clear those conceptual misunderstandings. [It works well] for tutoring, because they can look at problems, they can respond very quickly. Because of the data analysis and processing capabilities, they can deal with a very large solution base to get to appropriate personalized kinds of tutoring. And that's good.

I'm very bullish on the use of AI as a partner in the instructional process because I get tired of this conversation, "is it me or is it AI?" It's me and AI. And we can think about how do we judiciously employ, with our eyes wide open, because there are issues. How do we judiciously employ the capabilities that it provides for things that we do well that we can do better, [and] more importantly for things we have not been able to do? There are issues around plagiarism, all this stuff with ChatGPT, but there's also the issue that it provides simpler micro worlds for people to learn. If I can point a generative AI program like ChatGPT to the right sets of content resources and if I can supervise it in some way so that people can actually have simpler incremental pathways to learn particular things, that would be a good thing.

The UOC was doing open learning, online learning, you know, like I said yesterday before it got fashionable. And MOOCs and things like that. Many of our institutions, what we did in the first blush of online learning, we took what we were doing on site and planted it online, which is suboptimization and could be a bad thing also. We are missing the opportunity of rethinking courses and providing different kinds of experiences.

So, how do you provide help scaffolding and support to a learner population which is getting more and more autonomous in its learning? We cannot meet all their needs. We have to create facilities for them to learn on their own. Because right now the kinds of help and support we provide them are based on what we know from face-to-face experiences. Meanwhile, we have millions of users who are using online experiences. We have click-stream data, we have page-turning data, we have, we can identify data on how they find groups. So, we can come up at a macro level by using AI to mine the data to see how can we provide better support for online learners. I think that's the kind of potential we have.

When we were talking about AI, you said that there is probably a connection with authorship and plagiarism in assignments. What do you think the main challenges in this regard are? And what role do ethics and morals play in relation to all this?

It behooves us to make reasonable guarantees that they're not duped. So the data that they point to, the supervision…, that's a responsibility we have taken on as educators. That's a covenant practice.

And that's related to the quality and reputation of universities and institutions.

Absolutely. You mentioned ethics and morals. There was a colleague of mine who would say that you cannot legislate morality. But I think you also don't want to make it easy for people to confuse unethical practice with ethical practice. You want to have sometimes hard lines, but sometimes difficult thresholds so that they realize. Then we can go into some philosophical discussions about what is the truth. But I think we have to be careful. Even if you just treat it at the level of equity and fairness, you want to make sure that people who are striving to get to the truth are not distracted because somebody is getting it in a devious manner.

And then there's the risk of the bias of data. If you look at all the image programs, DALL·E and ChatGPT 4, it getting more and more sophisticated. My wife is an AI computer scientist. We have discussions. As educators, there are risks that we have to be mindful of. But there are opportunities we should not disregard.

Let's talk about lifelong learning. The demands of the job market mean that professionals need to upskill and reskill. They need lifelong learning to stay up to date. Here in the European Union, recommendations on the approach to microcredentials, lifelong learning and employability were made in June 2022. And MIT had the MicroMasters programs. What role will microcredentials play in the medium term? What needs to be done in order for them to be recognized by institutions or companies?

First of all, are they valuable? Incredibly so. For all of us here, there's a bunch of issues or considerations in your question. Are microcredentials valuable? How can you test their use? And then there are digital credentials. Because we are also working on that at MIT. When we first started doing open, and I'm talking about the early 2000s, when we talked about the value of open, we said boundarylessness. You can take courses from anywhere. Courses travel across the ecosystem, industry, education. It makes that boundary blurred, right? Was it you who told me that people are allowing industry offered courses and giving credentials at their institution?

Industrial doctoral programs.

Yeah. We are working with Latvia on this. But training programs that industry offers [and] universities are giving credits for that – that bidirectionality is a very nice thing. But we also talk about modularity. The project that we did here at the UOC with OKI (Open Knowledge Initiative) was saying, how can we take content applications and move it from institution to institution? If you can attach credentials to the modularity, why not?

People want some indicator of completion, and the marketplace, when it's hiring people, wants that stamp of approval or accountability. And if you have microcredentials, there are two advantages, and we're seeing this.

And companies are taking on a new role.

They're recognizing this. We want the competencies with these kinds of arrangements. I think what we are doing is create; let's say, if a student is a customer, we are giving the learner many more options and combinatorials to compose their experience. You mentioned lifelong learning, and that is really critical [for] displaced learners. Displaced learners are not just people who are geographically displaced, but also people who are vocationally displaced. "I'm in a job, I'm stuck, I can't leave, and I don't have enough time for a whole semester, but I can go in the evenings." That's displacement. "I'm vocationally displaced, and I can take these modular courses, I can get credentials for it, I can do it over time."

Some of the challenges we had in the old model of multi-campus universities – courses even won't transfer within the campuses in the same university – and we have come to a different point. Now, we have a different kind of challenge. How does a module from here with this credential interoperate with that credential over there? There's a lot of work going on with credential interoperability.