Reducing the cost of failure.

From Are Online Courses Failing America? by Megan McArdle. McArdle makes the critical point that MOOCs (Massive Open Online Courses) can't be measured by the same criteria as traditional courses. MOOCs, like many disruptive technologies, are not only substituting one thing for another (classroom education for virtual education), they are actually offering something different (greater access rather than limited access) and therefore have to be measured somewhat differently.

Commenting on high non-completion rates at one university:

The question is not whether they can replicate the current experience of going to college. The question is whether they can make it easier to get educated.

Start with the student population that was using the Udacity courses; many of them were high school students or in the military. These people were not substituting a MOOC for sitting in a college classroom; they were substituting them for not taking the class at all. Even if only 12 percent of students passed one of the classes, that represents a substantial number of people who might otherwise never have learned the material at all.

She has an additional point that is slightly more subtle but very pertinent to some work I am producing.

MOOCs will always have a very high failure rate. But that’s OK, as long as the cost of trying is low. Don’t have time for class right now? Drop out and come back when you have more time. Didn’t master Taylor Polynomials this time around? Do the course again.

As I’ll talk about in my forthcoming book, failure is often the best way to learn. More tries and more failures are almost always better than fewer tries and a lower failure rate. Letting people try a bunch of stuff, and fail at a lot of it, and then try again, is what makes the U.S. so innovative. We should welcome the ability to try this approach in education.

In decision-making in an uncertain and volatile environment, the priority is to make better decisions faster but there is an ultimate limit to decision accuracy. As you approach that limit, the alternate objective of risk mitigation becomes critical: how do you anticipate and mitigate failure so that you can recover quickly at the least cost possible. In making better decisions (increasing the probability of being successful), can you also reduce the cost of failure?