Personalized Video Learning

Dan Meyer just put out a nice blog post pushing back against some promotion of “Personalized Learning.” You can find the article here:

http://blog.mrmeyer.com/2017/problems-with-personalized-learning/

He’s not wrong- the applications the original article talks about are basic and only interesting if you compare them to pretty terrible teaching techniques. I’m not here to pile on about it, because there’s no need and Dan nicely covers the bases. I’m here because his article triggered an idea, and I needed a place to work that out.

In his response, Dan brings up the oft-touted benefit of video lectures: that students can rewind and review parts of the lectures to “further” their understanding, or to “clarify” difficult concepts. He also (rightfully) points out that that’s not how this really works- you don’t ask someone to repeat exactly what they said again when you miss something the first time. You ask for the variation- the alternative take. Another angle on the whole thing, that might better clarify that difficult concept.

But what if we could?

I’m not sure of the specifics of implementing something like this, but it seems like it might work like this:

  1. An instructor does a short video about a difficult concept. For that video, we follow current best practices and keep the video sub-six minutes, we have good graphics, speedy and natural delivery, and all that goodness.
  2. The same instructor records several other takes- and that’s not variations on the first video, but alternate methods. Maybe a metaphor. A real-world example. An animated version. Whatever.
  3. The student is eventually served that first video, and at the end of viewing, is presented with something like a “Got it?” button. If yes, move on to the next thing. If no, student gets served one of the alternate videos on the subject. Repeat until the answer is yes.
  4. Over time, analytics show which video versions are most useful (as they’re likely the last version watched…), and we can start feeding that data back into the creation of video segments (so, for example, if the animated graphic version seems to do best, that becomes the primary video in the future).
  5. Indeed, most LMS systems would allow those video analytics to be collected per-student, and you could eventually begin to serve each student individually the version of the video most (statistically) likely to suit them. You’d want to keep that model somewhat fluid, as students will find different variations work best/better on different subjects.

I’ve not seen this done, but in retrospect it seems pretty obvious. If you’ve seen this work somewhere, could you send me a link? I’d like in on that action.