Instead of learning statistics from a robotic professor, learn it from an actual robot!

by Andrew Gelman on May 23, 2012 · 5 comments

in Education

Steve Kolowich reports:

In experiments at six public universities, students assigned randomly to statistics courses that relied heavily on “machine-guided learning” software—with reduced face time with instructors—did just as well, in less time, as their counterparts in traditional, instructor-centric versions of the courses. . . .

A total of 605 students were randomly assigned to take the course in a “hybrid” format: they met in person with their instructors for one hour a week; otherwise, they worked through lessons and exercises using an artificially intelligent learning platform developed by learning scientists at Carnegie Mellon University’s Open Learning Initiative.
Researchers compared these students against their peers in the traditional-format courses, for which students met with a live instructor for three hours per week, using several measuring sticks: whether they passed the course, their performance on a standardized test (the Comprehensive Assessment of Statistics), and the final exam for the course, which was the same for both sections of the course at each of the universities.

The study was conducted by William Bowen, Matthew Chingos, Kelly Lack, and Thomas Nygren.

I don’t have much to say about this. The results make sense to me: as long as students are willing to do the work, I’d expect that an online system would offer better feedback than the traditional one-size-fits-all homework assignments.

But I think this is a bit silly:

Engaging students, such as professors might by sprinkling their lectures with personal anecdotes and entertaining asides, remains one area where humans have the upper hand.

Sure, I’ve been known to share the occasional personal anecdote. But I think the real value of the professor is his or her depth of understanding of the material.

Also, I don’t understand the claims about the economics. Kolowich writes:

In terms of instructor compensation, the researchers estimated, a machine-guided course featuring weekly face-to-face sessions with part-time instructors would cost between 36 and 57 percent less than a traditional course in which a full professor presides over each 40-student section.

I’d think the savings would be much more. You’re comparing 3 hours a week of a full professor, to one hour a week of a part-time instructor. If the prof gets paid three times what the instructor gets per hour, then shouldn’t the cost savings be 89%? OK, not exactly, someone has to set up the computers etc., so maybe you would get cost savings of only 70% or 80%. But how do you get down to only 36-57% savings??

In any case, I think this sounds like a great idea. My challenge in trying to put together something like this for intro statistics classes (the programmed-learning-and-computer-feedback part, not the get-rid-of-the-instructor part) is that I’m still torn about how to structure the course material itself. I don’t really like the idea of giving students 14 weeks of confidence intervals, paired t-tests, goofy probability problems, and the like. I’m still struggling to organize something that makes sense.

Conditional on the material that’s covered in the course, though (and without tying it to getting rid of the professor) I think programmed learning is great.


Comments on this entry are closed.

Previous post:

Next post: