The sudden rise of the Massive Open Online Course (MOOC) has been a shock for many in the education industry - no more so than to programmers engaged in creating educational software.
The shock is that the methods used by these hugely successful courses are little changed from the dark ages. If you think I'm being a critical outsider, then you might like to know that one of the leading lights in the movement, Peter Norvig, agrees with me. He even makes a joke in his recent Ted Talk that lectures in MOOCs are just like the 16th century monastic lecture theaters complete with the guy at the back sleeping!
The trouble is that unless you realize how things could be, you just don't see the lack of progress in the current crop of courses. I was talking with a colleague who used to create educational software about MOOCs and phenomena like Khan Academy and we agreed very quickly:-
"if we knew it was that easy we both would have done it years and years ago."
Instead we tried to do difficult things that still haven't been done.
The Video Lecture
If you look at any of the examples of MOOCs today then what you will see is a traditional lecture transposed to video. Most of the time there isn't even any attempt to clean up the presentation - not even a carefully drawn diagram and don't expect to see any animations. This is the cult of personality in action. The lecturer is center stage and they either carry the student along with them or they don't on the force of their off-the-cuff explanation and charisma.
Some can lecture, some think they can lecture and other are just terrible at it.
Overall it has more to do with acting ability that it does with say understanding of the process of education. The best lecturers have a presence of a maestro and understand their subject well enough to know how and when to say helpful things.
You can justify this in that when it works, it works well and when if fails very little has been lost - there's another lecturer along in a minute.
However, we should not pretend that is is optimal even when it works.
Recent surveys of how students respond to the style of presentation has revealed that they prefer hand drawn diagrams and unscripted lectures to carefully worked out pre-prepared slides. This is understandable. Having lecturers engage with their material by interacting with their own presentation is obviously superior to them simply presenting a static slide and talking around it. This is what gives a PowerPoint, or any slide based presentation, a reputation for being boring. The carefully prepared slides are dissociated from the lecturer's presentation and at best they are used as a crutch to keep a poor lecturer on topic and making progress. In a traditional lecturing situation, of course, spontaneous presentation has a good chance of being better.
Again when spontaneity works, it works reasonably well and if it doesn't, it was cheap,
However, we should not think that this is the best of all possible worlds - it isn't.
Back in the 1990s there was a trend to create lots of edutainment software. The software used the PC to present ideas, provide games with educational goals and generally use every facility that a computer had to keep the student involved, motivated and learning.
Some of the titles were very good indeed, but such titles were expensive to produce and schools and colleges had little money to buy them and little incentive to use them. For one reason or another this golden age of educational software is now just a memory - or it would be if very many people bothered to remember it. What is surprising is how few references there are to it or what it was trying to achieve.
Even by the standards of the edutainment software of the 1990s, the methodology of the average MOOC is primitive.
Let's consider what a computer can do for us.
Diagnostic Driven Non-Linear
The first and most obvious thing is that learners are individuals. They want to work at their own pace and they need to take different routes through the material.
Surely no one can argue with this point and yet we have MOOCs that present a linear flow of lectures, with tests and quizzes leading from one to another.
Is this is the best we can do? The software is quite capable of taking the test results and using them as diagnostics and working out what the students haven't managed to learn. It could also then select the material that would help the student the most.
Of course the MOOC organizers make a lot of the fact that it is the community that mentors the student with problems. This sort of works in the same sense that the rest of the system sort of works.
When it works it works well, but when it doesn't ....
The community is a powerful force but how useful it is depends on too many arbitrary and accidental factors. Whether or not a student gets the right help depends on how they ask the question and how lucky they are to engage the crowd's interest.
The only real advantage that crowd mentoring has is that it is cheap and perhaps it carries with it an overtones of egalitarianism.
Learning shouldn't be based on an ability to interest the crowd and learning outcomes shouldn't be left to chance.
The point is that a path through any topic should allow for non-linear progress. There have been many attempts to automate education that have created graphs of a subject and how it can be navigated. None of this is new and yet we are still relying on the linear method of the dark ages.
It isn't just the diagnostic and guidance aspect of the software that we are missing out on.
Modern software with 3D graphics, sound and interaction is capable of showing the student how things work in ways that an illegible free hand drawing cannot.
For example, if you want to show how an internal combustion engine works you can adopt the MOOC approach and get the lecturer to draw diagrams explaining as they create each sketch. This sometimes works and it is cheap.
The alternative is to provide a cinematic presentation - a simulation of the engine, complete with zooms and pans, and provide a commentary explaining what is going on. It should show the basic principles and then go on to show the subtle points - the way the gas mixture burns, the forces on the piston and so on. It should guide the student and it should also be interactive.
You might say that this probably works for a topic which is concrete, but for abstract things like programming, probability or whatever it just isn't possible. The only reason you might think it wasn't possible is because we don't even try any more - because high quality graphics are expensive and we can always sketch something that looks dynamic and raw.
If there is an idea inside the expert's head of how something works, programming or probability say, then there are graphics, usually animated graphics, that can present the core of the idea better than simply talking about it ever could. The problem is that it can be difficult to get right and this makes it expensive.
You can transfer the abstract understandings inside an experts head much more effectively using interactive simulations of the fundamental mechanisms than by using only words. Once transferred, the words that surround it help make the whole thing clear and asking a few questions that are difficult if you don't understand and easy if you do provides the diagnostic.
You can see that there is a repeating pattern here.
Dark ages methods are cheap and when they work they work reasonably well.
When they fail, who cares, there's another student and another lecturer along any time now.
The alternative "modern" methods are not cheap. They require lots of well developed software and the knowledge of the subject expert to be melded together to produce something effective. When it is effective then it should work for a wide range of students and even auto-detect the few special cases where it doesn't work. Once the materials have been developed, the methods are easy to apply and easy to replicate and makes best use of students and lecturers.
As programmers, our challenge is to create the tools for the job. Interestingly the tools are the same ones we already use to create games. They only require to be augmented with some smart diagnostics.
M. James, "Computers and routes to learning", Bulletin of The British Psychological Society, 32 (1979), p. 420.