What’s Next for MOOCs
Hi, I’m Stephen Downes videoing in from Casselman in Ontario, Canada. My talk today is called What’s Next for MOOCs.
Now that might be a bit of a surprise for you because you probably read last year or maybe the year before or five years ago, that moocs are dead. Well, I’m here to tell you, first of all, MOOCs are not dead. In fact, we seen in 2020 that demand for MOOCs has risen considerably. Here’s something from Class Central looking at 2020 web traffic on MOOCs and you can see it went, for Coursera, from 27 million to 74 million. That’s a lot. If you look at the number of registered students in MOOCs, we’ve seen it go up quite a bit as well.
If we look at also the different technologies, the different platforms that people are using for moocs, we’re seeing not only Coursera, Udacity, Edx, the ones that everyone knows, but also MOOCs from the Arab world, such as Rawq, Spanish MOOCs, French Université Numérique, and more. So no, MOOCs are not dead.
What changed? Well (as you saw quickly on that slide) the discussion in the media changed, and went down from the peak from a couple of years ago or early in the decade to a dull roar, shall we say, but look at the chart on the right and we see that the positive green sentiment discussion over the last few years has gone up steadily as well. So no MOOCs are not dead. People still like MOOCs, they still use MOOCs and this year, especially, there’s been a lot about MOOCs in the air.
So that’s what this talk is about. Overall the contents for this talk: well, first of all, there’s the bit about MOOCs are not dead yet, but I want to spend more time looking at how MOOCs provide access to what we might call a global set of micro credentials and skills based learning, and arguably employers want that, certainly I think they want that. We also want to address the question of how MOOCs are improving, or maybe more accurately, how MOOCs could improve in terms of student engagement and completion. So we’ll talk about that a little bit near the end of this presentation.
So you’ve all seen the standard diagram, I’m sure, that says what a MOOC is. It’s a massive open online course. Some people are saying any of those terms are negotiable. Me, I think ‘open’ is non-negotiable. ‘Online’ is non-negotiable. ‘Massive’, well, we could quibble about that. ‘Course’, we could quibble about that, but basically, ‘open’ and ‘online’ are what make a MOOC. There’s also two types of MOOCs, the xMOOC and the cMOOC. The xMOOC focuses on content and the cMOOC – the connectivist MOOC – focuses on community.
Now, the big thing that has happened in the world that made MOOCs possible aside from the kind of decentralized approach that we used in the cMOOC was the scalable web application. What this means is (and here we’re seeing an example from Microsoft Azure and you don’t need to worry about what all the little boxes are) basically is it’s an architecture that allows us to use the cloud in order to add new resources as we need them. So if we have one to ten people looking at our MOOC, a very small MOOC, then we have just a few resources. But if we have a hundred thousand people looking at our MOOC, we use the cloud and we scale that considerably to allow for more resources, more bandwidth, more disk space, in order to make those MOOCs possible.
Now, this is moving into the domain of what we call serverless Computing or a serverless web application. And this is what I’m trying to do with gRSShopper, my own MOOC application. And the idea here is that the MOOC is just a single page web application and then all the resources, all of the functionality, are cloud-based. So there’s a content distribution network (CDN) for static content. And then there’s an application programming interface (API) that’s connecting the two things like, say, my database directory and my ‘Active’ directory, in the case of Microsoft, or if we’re using Amazon, it might be my IAM directory (Identity and Access Management), (or it might be) different kinds of resources (such as) authentication, artificial intelligence, content algorithms that produce graphics, even things like chat functionality, etc.
All those weren’t really a part of the early moocs (but they’re becoming a part of MOOCs today). So, okay, where do we go from here? Well, what we have is as these MOOCs have become more common and more widespread and and more cloud-based. We have continuous integration and delivery. There’s a whole set of tools that web developers use to build, ship and run these MOOCs. Basically, they automate every step of the process. This is the key to making content scalable, whether its large content or micro-content, in fact, especially if it’s micro-content. (So we) automate the process of designing the page, serving the page, setting up the software, setting it up for a particular user. All of this is done using what are called ‘DevOps’ (a whole subject in itself).
So moving on from there what we’re looking at are the additions of automated services to do the things that take us so long to do now in education. A good example is automated grading. Now, you’ve probably seen automated graders for multiple-choice tests. They’ve been around for a while, but there are automated grading systems that today are looking at short answers, essay length answers, and even automated assessment systems for skills and practices. (For examplre) we worked with a device called NeuroTouch which in which a doctor practiced doing neurosurgery and the camera would look at the neurosurgery and make the determination whether we did it correctly or not. That kind of automated marking is coming and that is needed especially if we’re going to offer many many micro-contents and micro-credentials.
So, okay, what else then? Well, we need to keep track of all of this and we need to do this automatically and we need to keep track of systems of MOOCs from multiple different systems. So the mechanism that’s been devised for this is called activity records, and here you see a sample activity log (again, little tiny text, don’t worry about what it says). (Though, if you’ll note, the text that’s turquoise or green or something: for those of you who have been following my other presentations, you’ll recognize that as a hash address. So these are in many senses content-addressed learning records, but that’s an aside.)
The main thing is, whether you’re working on this learning management system or that application or that quiz system or even on things like remote devices sensor based devices, these records are captured from these different sources (and) gathered into the learning record store (LRS) for the purpose, for example, of things like automated assessment and also for the provision of new resources based on what you’ve done in the past.
So okay we’re seeing so far is pretty automated scaled up kind of process. So where do we go from there? Well, now what we’re looking at is actually organizing the content. We organize the content typically by credentials. These credentials, or micro-credentials, are taken to represent competencies or skills. Perhaps learning outcomes. But the main idea is that instead of getting one credential at the end of the course, we have for each of these minor achievements an individual credential.
You’re probably familiar with things like ‘open badges’, where you get a badge for this or that. now the thing with these micro credentials is the idea is that they’re supposed to be ‘stackable’. And so what we do is we take this micro content and we set it up in terms of different learning paths through a body of knowledge. The system may make recommendations as to the best learning path, or if it’s more an open-ended, it’ll allow you as a learner to choose the learning path that you want, and then as you complete each element in the learning path you’re awarded the badge. So this badge is the recognition of your progress through these learning paths, and again all of this is done automatically, so now we have an automated system for for recognition and an automated system for content delivery, all based around these bits of micro-content.
Now you might be asking about this content. Where does it come from? Because it takes a lot of time to produce content. Well again, automated content generation comes to the rescue. You may have heard recently of an artificial intelligence system called GPT-3. It only came out this year and it has been producing all kinds of automated or AI-generated content. If you look for, you know, “this face is not a real person” or “this is not a real dog” or even things like deepfakes, etc., this is content that was created by GPT-3 from scratch. It was assembled and created by it. So it can produce images. It can produce poems. It can produce songs (some very bad songs; there’s a channel out there on the internet “all heavy metal” produced by artificial intelligence. We don’t want this but we have it).
So there’s no reason not to believe, and in fact every reason to believe, that our micro-learning content is going to be produced more and more by artificial intelligence. Now, that could be a bad thing, right? Yeah. We’re seeing it already on things like YouTube or automated content producers for web blogs, etc., where we’re getting this this dump of content. It is supposed to look like user-generated content (but) you get these videos on YouTube (and) the voice sounds a bit funny. It’s like, “we’re doing the top 10 best movies of 2019” and they’re not really the top ten best movies of 2019, it’s an algorithm has generated this video, generated the voice, and you’re sitting there watching it not even realizing that its content created by artificial intelligence.
The way these systems work is they need to get their data from somewhere. Now an automated content writing system like the one that’s used by The Washington Post to write Sports stories gets its information from the event itself. Something like baseball, especially, everything is logged, right? Everything is recorded, every pitch, every hit, every error. Other sports (such as) hockey, basketball, etc., again, they all have their notation systems for what happened when, so it’s fairly straightforward for an automated content generating system to produce this content.
For other content, it depends on users. And what it does is, it looks at who said what when, maybe on Twitter or on Facebook, Instagram, whatever. It assembles it together and that’s how it produces its resource. So that automatically generated content is being introduced into things like WordPress (and) is being introduced into things like learning management systems. That day is coming very shortly.
So that’s where the micro content comes from, right? That’s where we get the the core, if you will, of our massive online courses. How do we make people stick to them? Well, as we can see on the slide here, the time that a presentation takes has a lot to do with whether people watch it. This presentation for example is only 20 minutes, which means we might keep about half of you all the way through, in theory, right? You know, I mean, it depends on the presentation, depends on the presenter, depends on the content. It’s not a straight line relationship. Nonetheless making these things shorter improves engagement, and I think that’s going to be the approach for massive open online courses.
Here’s one example (and I looked quite a bit for examples there aren’t many available yet, so this is still new): creating micro-MOOCs. Micro massive open online courses. So here’s a framework that was produced back in 2015. Here’s a slightly different framework, and one I like a bit better, (because) the micro contents aren’t all smooshed together into a single MOOC, but rather each one of these micro-contents is thought of as a separate independent MOOC and so as a result what we’re going to get is kind of a different approach.
If you look at this slide here, we see the difference between the old way of thinking of this, and the new way of thinking of this. The old way of thinking of this which was based on the idea of learning objects or open educational resources (OER). We had to aggregate or put this content together. You know, that’s where the idea of stackable credentials comes from, stackable course content, etc.
The new way of doing it what we have is a collection of web (or network – whatever you want to call it) of related content, but they don’t combine to form a single Unity. There aren’t set paths through it, but there are paths that might be defined implicitly as people in real time take one then take another and then take another, actually creating cow paths, which will be used by automated systems eventually to to recommend (learning paths).
So that’s the model of micro learning, right? So each one of these micro-MOOCs is a standalone course. A short course might be 10 minutes. Probably it’s going to be, you know, if you want to actually learn something from it, it might be as long as an hour. There are different models of these short courses. Some are just straight content delivery, others are based on dialogue or interaction, others are based on creating products of some sort.
So these courses form, if you will, a collection of community resource networks that are interrelated in different ways. They’re created by different communities. There are different paths through them, different producers of them, different consumers of them. So it’s an open-ended structure rather than a rigid structure. And the idea here is then people can take these courses as they wish. They might be offered through a learning management system. They might be offered and just taken directly from wherever they are by the student. They might be collected by a third party. It doesn’t matter.
Now just to make this interesting: we don’t need to centralize this system. And in fact what’s happening now. If we look at this slide, we have a decentralized fediverse made up of different ways of connecting these resources: ActivityPub, OStatus, DFRN. These are different ways of creating links from one resource to another resource to another resource, from one person to another person to another person. So (we have) individual entities in this fediverse, standalone entities, (and) these are microMOOCs, and then we have implicit pathways between them and it’s kind of like (if you read the quote on the slide there) it’s kind of like going into each other’s places and learning from them directly as individuals.
Thank you very much.
Oh, please do add your questions for Stephen. I’m going to kick off with kind of a broad question for Stephen. Just knowing his history, Stephen is one of the originators of MOOCs, and I wanted to ask Stephen, you’ve seen a lot happen since probably what about 2008, did you ever think it would get this big, that there would be that many universities and other providers involved in MOOCs, where we’re talking about getting credit for a MOOC now, did you ever think that it would get to this point?
I wouldn’t say I thought that MOOCs per se would get to this point, but I always thought that open online learning would become this popular. You know, even when we did MOOCs in 2008, all of us that were involved being working on MOOCs or working on open online learning for many years, a decade or more, so to me, I just saw it as one step in the longer process, and it didn’t matter whether this step is the one that took off, but open online learning, you know, given the affordances of the internet, open online learning is was and will be, you know, very large in the future and that’s what I expected at the time.
What did you expect at the time Stephen?
I think we expected 20 or 30 people to join us in our small little course where we described by demonstrating the principles of connectivism. It was like, “and then we’d all go on to our next projects.” I that’s what I expected because that’s mostly the results that I get.
But that’s interesting. Thanks very much for that. Stephen. One of our colleagues online just typed it. I comment he says classic @Downes. So I think your colleagues really, you know, know a lot of your history with with MOOCs. Our next question. How do you see the acceptance of the MOOC micro-credential building? Will schools and companies begin to see these credentials as valid?
That’s going to vary a lot depending on the credential. If you do and pass a course and receive something like say Microsoft certification or CGA certification, and these courses are of various lengths, that credential’s already accepted. If you do one from Slim Jim’s Course Building Technology R Us, that’s probably not going to be accepted so much.
And you know, I think we’re going to be looking at a model where different vendors offering different credentials (and I say vendors deliberately not because I’m pro-commercialization, but because that’s almost the way we have to think of them) different vendors offering different credentials will have different reputations.
Also in the longer term, past the next 5-10 years, we’re not going to need credentials so much as directly accessible proofs of competencies or skills. And that’s where a lot of this automation comes in. It looks at what you as an individual have made available online as a portfolio or as a trace of your online activity or as those learning records that I talked about, it looks at all of that, and it can be accessed by a potential employer or a potential contractor in real time and will deliver, I don’t want to say a verdict, but a recommendation saying, basically, “yes, this person is qualified to do the job that you’re asking them to do” or No this person is not.”
Now (there are) all kinds of ethical issues with that. I get that. But there’s all kinds of potential issues with credentials as well. And this gets rid of that intermediate step and gives us direct access to skills and competencies and that will be accepted in the future. It won’t matter where you learn them. It’ll matter how well you demonstrated them.
Thanks, Stephen. Our next question. What is the next barrier to be overcome by MOOCs, for example, do we really need professors if artificial intelligence can produce content?
You know, I thought about, I thought about that a lot, you know, because you know, I’m back in the days of learning objects. I remember those days the line was (and I know because I wrote that line) that the math lesson has been taught, the same math lesson has been taught since time immemorial. It hasn’t changed. We don’t need 14 million different iterations on it. And that’s still true. One single good math lesson will probably do the job.
But I’ve also learned over time as I think the rest of us have, especially this last year, being live is important. Being live allows for that quick back and forth exchange that that allows for real-time non-predictable responses to inquiries and banter, like what we’re doing now, and it’s important not just for the learner and I think this is really key, it’s important for the provider (in this case me) because that’s where a lot of the new ideas come from, and that’s true on both sides of the interaction.
So I think, you know, we don’t replace everything with automation. We got a lot of automatically generated content for sure and the idea of the role of the professor as the content delivery mechanism fades because there are better ways to do it, more efficient ways to do it, but the role – actually going back to even the traditional role of the professor as someone to bounce ideas off of someone to engage ideas and brainstorming or speculation or trying something out that remains imprecise -that’s still a valuable thing. Long long term, maybe machines do that too, probably machines do too, 50 years from now machines will be able to do that too. But in the interim between now and 50 years from now, we’re still going to need a human to do it.
It’s good to know Stephen has really appreciate where I’m not going to get eliminated in the next few years. That’s great news.
Well depends on your job, right your job here as host…. Sorry. Just kidding.
Thanks so much. If you had to pick out one thing that’s going to happen with MOOCs, say in the next year. What would it be next year?
People will be saying the pandemic didn’t need mean anything. We’re all going back to in-person education. MOOCs were overhyped, online learning was overhyped, we learned it was a horrible terrible experience. We should never do it. That’s what I predict will happen in the next year. I think a lot of that reaction will be over-reaction, and over-reaction is something that you can rely ably predict in any circumstance at any time.
But you know, I mean, we kind of come back to this, I like (if I may pat myself on the back) the idea of the micro-MOOC, because I’ve sketched out and I really only came up with that while preparing for this talk, so it’s a bit new to me and I haven’t really thought it through a lot, but it seems to me that producing a bunch of course sized micro MOOCs, and I have to think about exactly how they instantiate themselves, seems like a logical next step in what we’re doing.
I mean you certainly address the problem with completion, right? I mean, that’s that would that would do the job in itself almost, so I think we’ll be looking at shorter form MOOCs. They’ve been getting shorter and shorter. Our first MOOC was 8 weeks, the longest MOOC we did with Change 2011 was 30, I think, 31 or 35 weeks. We all agreed at the end that that had been a terrible idea. It was a great MOOC but it just went on and on. And now we’re seeing two-week, three-week long MOOCs. So (we get) the idea of a one-session one-event MOOC that takes an hour and then you’re done. You know, really all that needs is the technical framing for it for that to happen and that could happen sometime this year. I wouldn’t be surprised because I’ll try it. Yeah, but mostly we’ll see the negative reaction.
Thank you for that. Another question. Is there anything that MOOCs can do to help overcome the learning gaps that were expecting because of the interruptions from the covid-19 pandemic. Are they a solution to help us address those gaps?
The learning gaps thing is a hard thing now, and it’s a hard thing for a bunch of reasons.
First of all, I mean we had we have the situation where electronic media failed to fill the gaps while we were in the pandemic, (so we ask) electronic media fill the gaps after the pandemic? It seems hard to say yes to that when you put it that way, right? But I still want to say yes to that because electronic media, all MOOCs, accessed on a voluntary basis are probably better ways of accessing learning than electronic media where you have to do it or you fail. So I think that’s part of it, you know. And also, too, this whole crazy pandemic thing goes away and that makes everything a bit easier.
But the other side of this question, which is hard is the whole concept of gaps. If everybody stops, where’s the gap? Right again. Yeah, it’s not like a few people kept going and everybody else had to stop and the gap formed between them. Everybody stopped, or more accurately, everybody slowed down. So there isn’t a gap in that sense. There’s only a gap in the sense of we didn’t meet these predefined learning outcomes for these years in our educational career. So well, that’s not a problem either. You just shift the outcomes and you’ve addressed the gap.
You know things happen in people’s lives. And this is a thing that happened in everybody’s lives all at once which actually makes it a bit easier to deal with. You know, the only real sense in which there’s a gap is that you don’t know at the age of 15 exactly what we thought you would know at the age of 15. Instead what you know is what we thought you would know at the age of 14 and a half. Viewed that way, given the decades of life ahead for said 15 year-old, I don’t see an issue personally, and I know a lot of people will play it up as an issue, but (I don’t).
That’s the third thing about this whole thing. It’s not like people suddenly stopped learning for the last half year or more. It’s that they started learning different things. They didn’t study official curriculum, but they didn’t stop existing. So they learn something and that’s where the more interesting outcome from the pandemic educational experience will be. What did people learn that was off the curriculum instead of learning math 20, say? And that I want to see the answer to. I think there will be an answer. I think we’ll learn things about how we learn and what we learn when we’re not doing formal learning, but I don’t know what that is yet. I can guess but they’d only be guesses.
Did you want to guess?
We learned how to work together online more? Yeah, we all developed our communications networks with our family, with our friends, and we won’t abandon that. So in many ways many of us became more computer literate. I got Zoom up and running for my parents who are in their 80s. They didn’t successfully use it but there will be many cases where the same sort of thing happened where they did successfully use it, and if you know that, take that and multiply it a million, 10 million, a hundred million times, that’s part of the experience. So suddenly being capable of interacting pretty fluidly through digital media is one of the things
So, you know, now what comes out of that? I don’t know. I think interesting things came out of that (like) new forms of organization, new forms of government, governance, companies that do a lot more learning or a lot more of their work online rather than offline, which means they do a lot more of their training online rather than offline. I’m already seeing that with NRC. I’ve worked at home since March I’m not going back. You can’t make me go back. I think a lot of people will be in that mode too.
Great. Thank you for that Stephen. We’ve got a question: what types of technologies are needed to enable community resource networks. How do you see these networks evolving over the next few years to support personal learning?
Yeah. Okay. There’s a list number one you need identity. And now I know that seems like an odd place to start but I think to necessary place to start. One of the things we saw during the pandemic is that it’s really hard to work with 11 applications, 11 logins, 11 passwords, and then trying to sync all of that together. It’s hopeless. Now those of us who have been working on this kind of stuff for a long time already knew that, but now that knowledge has probably percolated all the way into executive suites and decision-makers offices, which means that we will get a lot more serious about identity, digital identity, as well, a lot more serious about how we protect it, or associate that identity with resources, how we protect the ownership and control of those resources with respect to that identity, et. That’s probably the big thing.
The second big thing: I showed you that diagram with all of the different protocols for the fediverse, getting that down from like 15 protocols to a few protocols or you know, just even just a Rosetta Stone for them would go a long way. ActivityPub is on the verge of being the dominant communication mechanism behind the Fediverse. So if we have identity, if we have ActivityPub resource sharing and if this is adopted in any way that will go a long way.
Third and finally, institutions have to want it. I know that’s not a technology, but that’s a real requirement. If every company and every institution wants to put up walls and do only stuff within its own boundaries, its own borders, then we don’t get (community resource networks). These institutions have to accept that data moves from institution to institution, and not just at the high-level “we’re sending money” kind of protocol which already exists, but at the low level “one person is sending something to another person” kind of level.
We need that, and I think they will, I think, as I say, you know, the decision makers may have finally realized that we need this as a result of this pandemic, but we need that kind of desire on the part of institutions to actually create something like this.
Thank you. Stephen. We’re just winding down. We got a couple minutes left Stephen and I think what I might do is just turn it over to you to share any last words of wisdom from Stephen Downes on what’s next for MOOCs before we wrap up the session, last words of wisdom.
So OK. Well this comes from nowhere.
I mean, first of all, there’s going to be the negativity there already is a lot of negativity about ed tech and electronic media at all of that and the concerns that are raised by people like Audrey Watters and others are valid and legitimate and they are problems that we have to address. They are not arguments against ed tech, digital media or massive online courses.
They are arguments against specific models, models where they do surveillance, big data gathering, artificial intelligence-enabled projection, where they do things to you, right? That’s the model that we really have to worry about.
Being based on open online resources, the other side of MOOCs has always been that someone’s education, your education, is what you make it and we want you to make it rather than us to make it and deliver it to you. All the different things that I talked about, including micro-MOOCs, today are just ways of making it better and easier for people to make their own education. Because it’s when people create these things for themselves using tools that are easy to use and intuitive and widely accessible, that’s when we avoid most of the danger scenarios that are posed by critics of ed tech surveillance and the rest.
So that would be my final thing to say. Education is not about us providing a service for people it’s about us providing ways for people to provide for their own needs. The more we see that as our objective, the more successful will be in the longer run.