Daphne Koller What We Are Learning From Online Education
Written by Alex Armstrong
Saturday, 04 August 2012
A recent TED talk by Daphne Koller suggests that we have much to learn from online courses, but it might not just be the knowledge that they convey.
Coursera claims to have 640K students in 190 countries, but it has had 1.5 million enrollments. This is not a good drop out ratio - but why does it matter the courses are free and students sign up on a speculative basis? If the course doesn't work out, then students just give up. It is so casual that it hardly constitutes dropping out.
In this TED talk, Coursera founder Daphne Koller puts forward the view that online education can be made better by the huge amount of data being collected on how students perform. However, when you analyze what is being said the future doesn't look so bright.
The repeated theme is that top quality education comes from the top educational establishments. You hear the term "a Stanford Education" more than once. However, there is guarantee of quality of education delivery based on the standing of the university. In the real world, it is a self- fulfilling prophesy as good students gravitate to the best universities where the best researchers do research and occasionally teach. Whether they do the job of teaching well or badly really doesn't have much to do with the outcome - money, motivation and the fear of failure keep the student at the task no matter how terrible the lecturer. If you don't believe me take a sample of some of the videoed lecturers on You Tube - the range is remarkable.
Of course the aim is laudable - to bring free, high quality education to everyone. However, this isn't an excuse for not doing things well.
The "high quality" part isn't something that follows just because there is a Stanford or MIT label on the video. Currently the technologies being deployed are primitive. We have videos of talking heads with simple computer marked questions. Rarely are there any creative uses of software or diagnostics. What is worse, is that for all their mass market appeal the courses are flawed. Questions are ambiguous, a near correct answer is marked just as wrong as a something that was wide of the mark and so on.
There are some attempts to make things better. In the video we are offered the example of analyzing the frequencies of wrong answers to a multiple choice question. Clearly, if the majority of students pick the wrong answer then it is the course, or more likely the question that is at fault, not the students.
This isn't revolutionary - it's standard educational testing theory and it seems it has been rediscovered. The solution to the problem offered is a custom message to the students who got the question wrong - aiming to fix their misconceptions. This is a band-aid fix. The correct solution is to go back to the course material, or the question, and discover what was causing the misconception in the first place.
The main point to take from all of this is that lecturers are subject specialists. They are not necessarily skilled in teaching, and they might not even think particularly deeply about it. If you have had the opportunity to go to a physical university then you will know exactly what this means. Some lecturers are great and other are more interested in their research to be bothered by students.
This is a long standing problem and not one created by the new MOOC phenomenon. What is new is that the introduction of computers provides the first easy opportunity to attempt to bring the educational standards up to the very best and not simply because they have a well respected name stamped on them.
Udacity has opened a new Nanodegree in Machine Learning. At advanced level it prepares you to apply predictive models to massive data sets in fields like education, finance, healthcare or robotic [ ... ]