Can Online Education Lower Costs and Improve Quality?

EdSurge Podcast

Can Online Education Lower Costs and Improve Quality?

By Jeffrey R. Young     Feb 12, 2019

Can Online Education Lower Costs and Improve Quality?

This article is part of the guide: The EdSurge Podcast.

Inspired by the breakout podcast Serial, four years ago two digital learning leaders at the University of Central Florida created their own podcast—focused on online learning instead of true crime.

It’s called the Teaching Online Podcast, or TOPcast, and co-host Thomas Cavanagh says he is driven by his quest to figure out one of the grand challenges of higher education: how to use technology to raise the quality of instruction while lowering costs. Not everyone thinks that’s possible, of course, and even Cavanagh, vice provost for digital learning at the University of Central Florida, admits that edtech can spark plenty of new ethical challenges along the way.

Each month, he and co-host Kelvin Thompson executive director of the Center for Distributed Learning at UCF, give their analysis of trends in online learning over a cup of fancy coffee—and these days their fans often send them beans to brew and fuel the show.

EdSurge connected with Cavanagh (online of course) to talk about what he has learned from all those podcast chats, and about how his sidegig as a detective novelist shapes his work in campus innovation.

Listen to the discussion on this week’s EdSurge On Air podcast. You can follow the podcast on the Apple Podcast app, Spotify, Stitcher, Google Play Music or wherever you listen. Or read a portion of the interview below, lightly edited for clarity.

EdSurge: So what’s the mission of your podcast?

Cavanagh: The format is just a couple of colleagues sitting around talking about stuff that we do—which is online learning—over a cup of coffee, which has become kind of a recurring meme. But the idea is that if you are in this space, you could put on your headphones and grab a cup of coffee and sit down for 25 minutes or so and be part of a conversation with some colleagues. Occasionally, we do bring in interviews.

Your most recent episode was about learning analytics, and listening to it reminded me that the focus of edtech folks these days is less about the tools being used and more about finding ways to improve student retention and learning. Do you see that as a broader trend as well?


This week's podcast sponsor is AWS Educate: With a mission to provide students and educators with tools needed to accelerate cloud learning, AWS Educate will soon announce its newest offering for educators. Take a look at AWS Educate's evolution and hear about what is next.


I do, and I think as the tools improve, they will become more ubiquitous and more effective. But I am sort of all-in on analytics as a concept. I'm not sure if we've quite realized the vision yet, but let's leverage technology to help.

There are things that technology can do that humans can't. That was kind of the way in to the episode you referenced. We could probably do multiple episodes on learning analytics, maybe there's a whole podcast about it out there somewhere. But our way in was to kind of talk about the human side of it.

That means seeing analytics as a [supplement] to the human connection. That’s necessary when you're actually talking to students about their performance or their risk level or what they're predicted to do—whether or not they'll be retained in a major or at the university or succeed in a course. Those are really delicate conversations that probably should not be wholesale handed over to technology. But technology can play a part in it, and can provide information to the humans who are having those conversations. That was mostly what we talked about.

You're talking about enhancing human conversations with software robots. How does that work?

I guess it depends on the use case that you're talking about. For example, if you look at the Jill Watson experiment at Georgia Tech, where they created a virtual teaching assistant, the students in the class didn't know that they were talking to a bot. That is, actually, not a bad application of outsourcing a traditional human function to a robot because the service that it provided was worthwhile and helpful, and it did no harm. But if you're talking to a student, saying, “Hey, Jeff, it looks like you're not gonna be the aerospace engineer you think you are because of your performance in these classes,” then I don't think that's something we want to outsource to a bot. That's something that a human being needs to have a conversation with you about. But the human being would be informed by the data from some of these big-data systems, or AI or analytics.

Some critics of learning analytics worry that algorithms could end up reinforcing stereotypes and actually having a negative impact for some students. How do we avoid some of the pitfalls of getting algorithms into the classroom?

Yeah, it's a great question, and I'm not sure as an industry we've completely gotten our arms around it yet. But I would say maybe a place to start is: Don't prejudge any student; judge them on their actual behavior. We recently did some analysis, out of my department, where a faculty member who’s a world-class data scientist spent half a year analyzing some of our student information system data and our LMS data. Really what he came back with… is that incoming GPA on day one in the class is really the only thing that predicted success without any other data. It had nothing to do with race, ethnicity, age or gender. It was really just GPA, which was interesting.

So he claims an 85 percent accuracy of his prediction model before a student's done even one thing in the course. Now we need to validate that. But then once you start layering in LMS data, his predictive model goes up to a 90 percent accuracy, just based on a student's actual performance in the course. And again, none of the demographic data made a difference. The number of logins [to the learning system] didn't make a difference. Really what mattered where things that were graded, like quizzes and tests and assignments.

So, if you look at those two data points, GPA and actual performance in the course, and build your predictive models around those, you can avoid the bias of pre-judging a student based on any other kinds of factors.

Aren’t things like performance on tests and quizzes something that professors already know? Are you finding enough insights from learning analytics to make it worth the cost and trouble of setting up this huge infrastructure?

I think it's a question of scale. If you're teaching 12 students, then you kind of don't need the big data because you can have those kinds of deep interactions with your students and know what they're doing. But we've got 68,000 students here, and it's hard to treat each one of them individually on that level. I think technology could help us with that.

If we can leverage the AI or bots or data analysis to raise to the surface those students who are struggling and we could intervene quickly, I think technology could do it in a way that humans can't at the scale that we're talking about. And then even just beyond our institution, I think across the country, combining some of these data sets, learning about what's effective. I think that that's the promise—it's a promise of scale. If you and I were the only two people in the course, it would be easy.

What are some of the lessons you’ve learned from the interviews and conversations you’ve had on your podcast over the last few years?

I've learned how many people like coffee. Because, God bless them, they keep sending it to us to drink on our show, which is kind of cool. So we've got a long backlog of coffee to drink from generous listeners.

That's a pretty good perk.

If there's one thing we've learned, it's that people are interested. I think it's a reflection of the growing ubiquity and strategic importance of online and digital learning in various places across the country. It's hard to hire instructional designers now because they're in demand. All of that's a reflection of the kinds of things that I hear.

You're probably one of the only people we’ve talked to on the podcast who also writes murder mysteries. Is there anything that your kind of work in that space has taught you or made you think differently about your work in digital learning?

I don't know. Murder mysteries and online learning. I probably should firewall those if I haven't already [laughing].

I love doing both. I think, honestly, the creative outlet of fiction writing makes me better at work. I think it exercises certain kinds of critical thinking muscles that can be applied to everyday work, even if it's looking at a spreadsheet. It probably doesn't hurt to be thinking in different ways, solving problems.

I guess it's not like you're going to literally make a choose your own adventure online course, but it does strike me as something that is a pretty interesting parallel life you're leading.

Yeah. Honestly, I think if you get in the habit of writing, it also helps you with the scholarly writing that I occasionally do. I know that blank page can be intimidating, but if you just sort of get in the habit of writing all the time, whether it's fiction or whatever, [you can say to yourself,] “OK, I'm just gonna write this thing now today instead of that thing.”

Any other takeaways from all those coffee-fueled podcast conversations?

Well, there is a recurring theme that we keep coming back to in the podcast that I don't think is going away anytime soon. The idea relates to the “iron triangle” and how online digital learning is a potential disruptor for breaking it. The iron triangle is access, cost and quality. Each of them is a bar on the triangle, and the thinking is that you can't positively impact all three at the same time. So if you wanna increase quality, you're going to negatively increase cost, for example. Or you're going to reduce access.

The notion of disrupting that through online learning is that you can positively impact all of them by using different models and new ways of doing things. It has been our premise here. We're doing everything we can to try to positively impact cost, access and quality through digital technologies. It's why I go to work every day.

Can you give an example of what that looks like in practice?

For example, you can increase access and reduce cost through a transfer program. We're the largest transfer-receiving institution in the country, mostly through our direct connect to UCF program. This is kind of separate from online learning.

We've got a consortial agreement with six [community colleges] in our region where a student can declare their intent to come to UCF when they graduate with their associate’s degree. If they're part of this consortium, they're guaranteed admission to UCF. So the first two years are completed at a much lower cost, and we're able to open access for those students on a guaranteed basis. So it positively increases access, reduces cost and then I would say our quality is at least as good, or at least not compromised by doing that.

We do similar things online. Like using adaptive learning, for example, to personalize the experience for students at scale.

Online learning allows you to reduce cost if you're doing it through some economy scale. I argue that you can increase quality if you do it well. It requires effort and investment. You're opening [college] up to people who previously could not come.

I use the example of my mom, who was a nurse. She worked the overnight shift in the ICU, at a hospital. If she had wanted to go back to school. she couldn't go to class Monday, Wednesday, Friday at 10 o'clock in the morning. But if she had wanted to go back and get the next level of her education, online learning would've been a solution for her. We have an awful lot of students like that here, so that’s definitely access.

Learn more about EdSurge operations, ethics and policies here. Learn more about EdSurge supporters here.

Next Up

The EdSurge Podcast

More from EdSurge

Get our email newsletterSign me up
Keep up to date with our email newsletterSign me up