Harnessing AI's Powers For All

The explosion of interest and research in generative artificial intelligence (AI) has created plenty of excitement about what the technology can make possible. While large language models like ChatGPT have brought new interest to the field within the last year, researchers and educators have been exploring these technologies for more than a decade.

For the last 15 years Visiting Professor Seiji Isotani has focused much of his research on how people learn with interactive and intelligent educational technologies. He’s also studied the impacts gamification — the use of gameplay-like elements such as rewards or incentives to participate — has on learning experiences. Isotani, a professor of computer science and learning technology at the University of São Paulo, Brazil, has worked recently to develop AIED Unplugged (Artificial Intelligence in Education Unplugged), AI systems to help teachers bridge the digital divide.

“We are reaching students that don’t have access to this technology,” says Isotani, noting many don’t even have the internet. “And they still do not have access to the technology, but now they gain the benefit of the power of technology by a proxy, which is the teacher in this case. We provide all the power of this cool stuff that AI is providing to the teacher so the teacher can be better for the student.”

AIED Unplugged has designed applications that have impacted 266,000 students in Brazil, using AI software to aid teachers in evaluating student writing, leading to improvement in students’ skills. That work has since extended into countries like Mexico, Peru, and the Philippines, where Isotani and other researchers help develop policy to increase AI access in remote and low-income areas.

“In a society where everything is changing so fast, we need to be better at understanding the impact of technology and moving along with its advances. Because if we don’t do that, we are slowly going to be left behind,” Isotani says. “It’s important to understand and have the balance to use these technologies and test these technologies and improve our practices based on the evolution of knowledge.”

Below, Isotani shares the impact his research has had across the globe and what lay ahead as technology continues to provide opportunities to bridge the equity gap in education.

How would you describe the focus of your research?

I have two pillars in terms of research and impact. The first one is how we can create technologies using AI that support students the most, especially considering their needs, their skills, their knowledge, their culture, and any other factors that we can put on the table. The second one is that it doesn’t matter how personalized we can be to support students if they are not engaged, we need to support and motivate students to be engaged while they are learning.

These two topics somewhat converse and talk with each other. And so, what I’ve tried to do is ask how I can use these two topics to create or support policies that actually reach out to the most people possible.

Artificial intelligence — how to grapple with it, how best to use it — is a huge focus in education right now. How does your research into AI — which is often focused on the digital divide and issues of access — stand out?

I remember when I was in UNESCO in 2019 and we were discussing this divide. AI wasn’t this big thing that it is today, but the trend was there. My work tries to rethink how we can design AI technologies to reach underserved communities. Those that are usually left behind. The idea is if you’re a student that does not have access to the internet, or a computer, or a cell phone, can [you] receive any benefit from these AI technologies? Most people will say no.

I’ve seen so many reports from OECD, UNESCO, and all of these international organizations asking governments to invest more resources in infrastructure. I don’t believe investing more money in infrastructure is going to solve problems in education. What we need to do is stop and think: Can we be better at designing these technologies so we don’t actually need better infrastructure? We can use the current infrastructure.

With AIED Unplugged you’ve focused on using cell phones to access this tech because there’s a much higher user rate at lower income levels. But what do these AI technologies look like?

You can use generative AI to do some different tasks. In our case, we take a picture of the essays of students to improve their writing. What we’ve done is process that picture to understand where the text is, do segmentation to identify lines and words. And then the AI is able to provide analysis of this essay and give recommendations to teachers.

We compile this large number of essays and the AI identifies what students are missing, what they are doing right, and gives recommendations to teachers on how they can approach students. It empowers the teacher with knowledge so they can act in class in a better way.

And I think that’s the beauty of AI: It’s being used to empower ourselves to be better in what we do. You can think about teachers, you can think about students. How can you use this technology to help students be better learners? To have more capacity to learn and be more engaged with the learning process?

How does the idea of gamification factor in?

In AI, I’m trying to adapt or personalize student pathways to learn, using gamification elements to provide nudges or incentives so they can change their behavior. We are doing work to understand the impact of these strategies on student behavior and then on student learning. Sometimes, we try to understand learning itself. Not just the behavior of learning or the process itself.

In the case of gamification, I like to quote the founder of Duolingo, Luis von Ahn. If students are not interested or engaged, it doesn’t matter if you have the best pedagogical approach. They will not learn as much as if they were engaged and motivated to learn. So before thinking about anything else, we have to motivate them.

Gamification gives us tools to engage students in different ways. Many people only look at the extrinsic motivation of students using gaming elements. But we can also touch the intrinsic motivation of students using gaming elements. Which is interesting is how we design the interactions and the way gamification will affect students.

So much of the conversation about AI is how it can be used to circumvent learning, but your work argues that it can to actually enhance learning. How do you strike that balance with new and emerging technology?

That’s an interesting question because one thing many people say to us is “My teachers do not want to use technology in the classroom.” And they are being left behind because they are overburdened with work. The problem is that changing culture is difficult. And when we are thinking of adapting technology, it’s not about the technology, it’s about the people.

We need a full cultural transformation, and that's so hard. You can be any big tech company and develop this super technology, but it won’t solve problems because the problems are in the day-to-day interactions. That’s why when we work in these huge, country-wide policies, we need to think about how not to disrupt an environment and actually use that particular space to find benefits and embed it in teachers’ lives so they do their jobs better. Then people start to adopt technologies and new strategies. With these small changes, you can be in a much better place over time.

Do you think some of the hype around AI has masked the fact that we still know very little about this technology?

I agree, we are in the hype phase. For good and for bad. The perceived risk is stretched and the perceived benefit is, too. We actually don’t know where we are going right now, and I don’t necessarily think that’s a bad thing. It’s only bad when people bet on those stretches. And I regret to say that many policies are being designed considering the perceived risks of AI in a very bad sense. Thinking that AI is going to destroy the world in so many ways, so we have to contain it in a black box and only a few people can use it and you can control everything.

I don’t believe that’s the right approach for any technology design. You need to provide nudges and pathways so companies and researchers can actually work in a free space where they can innovate and actually support their society.