Skill-Building Heuristics

After I gave a talk at Ephemerisle, some months ago, I got some requests for a summary in writing. Here it finally is!

My intuitive model of learning spits out that I learn best when I’m “challenged”. I’ve come to believe that a feeling of accomplishment and a desire to “consolidate” what I’ve learned (by taking a nap or some other sort of break) are good signs. I’ve also come to believe that frustration and failure are suboptimal. Sure, it’s virtuous on my part to be able to learn from situations that aren’t designed for my learning experience (most of them), and that means learning strategies to cope with frustration and failure. But when I’m the one designing the learning experience, for myself or someone else, I now see it as my job to minimize frustration and failure in the learner.

Errorless learning is a thing, and it seems to me to be a good thing. (At least in certain cases. I was going to say it was great for people, like amnesiacs, who relied on implicit memory more since I had remembered reading that theory, but while looking up citations, I also found this, which says that for memory, errorless learning can trade off against effortful recall, which seems to be important. Interesting. For memory, I think spaced repetition is quite optimized.)

An unfortunate side effect of learning that does involve failure and frustration is that you can end up getting misplaced aggression in the learner. I’ve definitely observed this in my dog, my daughter, and myself, though not to any serious degree.

I think TAGteach is a pretty good methodology. Here are my main takeaways:

  • Work on one criterion at a time. (Practice either speed or accuracy, for example.)
  • After giving whatever longer instructions you want, describe what the learner should do in at most five word.
  • Easily verifiable criteria are the best. (“Arms above head” beats “hold your arms up high”.)
  • If the learner can’t do what you’re asking in three (or so) tries, make it easier!
  • If you can use a marker sound (like a clicker, or saying “Yes!”) to point out the exact moment the learner got it right, this can speed things up quite a bit.
  • If you’re increasingly the criteria in one area, expect performance in other areas to degrade. This is normal and expected and will go away once you get the thing you’re working on straight. (If you’re trying to jump straighter and you’re not jumping as high as you were, that’s fine.)
  • Being in a new location (or with new equipment etc.) is an increase in criteria.
  • To best get fluency, work on precision before speed.
  • For learners who want to learn the thing, getting the information that they’re learning it is potent reinforcement. But if anything is making the experience unpleasant, it can be nice to additional actual treats when they get stuff right.

Punishment is very problematic for building skills. It has advantages, and can be very effective, sometimes resulting in one-trial learning. It also seems to generalize well, but threat kills curiosity and experimentation, which I very much want to preserve in any learners I work with. Here’s a Karen Pryor quote from Reaching the Animal Mind:

Clicker trainers have a strong gut reaction against punishment; and the more experienced we are, the stronger the feeling. Well, thanks to Panksepp, we have a scientific rationale for why mixing correction and reinforcement is harmful rather than helpful to the learning. It’s not just a moral issue; it’s common sense. Correction or rebuke switches the learner from the hypothalamus and its SEEKING mode to the amygdala’s path of avoidance and fear.

And even punishment that seems mild to the one administering it can be very bad for relationships. 

I will emphasize that, since learning more about learning, it’s become increasingly obvious to me how much people, including me, tend to assume that teaching means telling people they’re wrong and “correcting” them. As a tutor, I’ve found it especially hard not to fall into this pattern. My narrative has historically been that I’m providing maximum value to people when I’m showing them something they don’t know yet, and that equates to giving them problems they can’t do yet. These days, I’m changing my approach to something friendlier and more success-based :-).

I’ve also largely moved away from “teaching” as a paradigm recently, since I’ve been working on grokking unschooling, and that’s something I’ll try to write more about as my thoughts feel ready to be put into blog posts.