Tuesday, January 11, 2011

Force breakthrough by forcing mistakes

A week or so ago I started to play Math Workout to rekindle my degenerated calculation skills. After a while I got stuck at a certain speed, making no mistakes anymore. I tried to push myself to go faster, but without making mistakes but that just didn't work. Then I tried a different approach: I forced myself to go fast, mistakes or no. After a couple of sessions, I went back to making no mistakes and to my surprise I was around 8% faster. I have a theory about why this is and I'm applying it to life without hesitation from now on, so read on if you want to share the fun.

<!-- more -->
What's Math Workout
I'll give you some background on Math Workout first, so you can understand the example. Math Workout is a game that let's you solve simple calculations and times you while you do so. The main goal is to be as fast as you can. For example, on easy level the hardest addition is 8+7 and the hardest multiplication is 4*6 (I'm considering 5*6 easier). Dead easy as you can see. The trick though is that you need to solve 20 of these problems in less than 23 seconds, and I can tell you that's tricky enough because you need to read the problem, come up with the answer and type it correctly in about a second.

What goes wrong if you don't make mistakes
Somehow I got into a rhythm. See the numbers, see the operator, answer pops up, type it. There are probably a few more steps that I was doing unconsciously, but the point is that I was taking a fixed amount of time for each step. Sometimes I found myself breaking the rhythm just before I punched the wrong number and then taking some fixed amount of time again to correct myself. Whenever that didn't happen my time would be between 26 and 27 seconds consistently. Basically I had put a few safety checks in (which I probably didn't need most of the time) that were slowing me down. Whenever I pushed myself, I would occasionally parse the wrong operator, type the wrong number, pop up the wrong answer and quickly revert to the strategy that worked. Once I noticed this, I knew I needed to break the pattern.

Forcing errors
I decided to change the pattern, and ignore the mistakes at first. I decided to ignore the operator check and assume that my brain would be able to pick up the numbers and the operator in one go. I made 5 errors out of 20 questions and decided that I was bad at picking the operator, but it must be working somehow, otherwise I would have seen more in the range of 10 mistakes. I gave myself positive feedback by thinking I was a smart cookie for figuring that out.  I repeated it a few times and saw the errors go down and the times with it, more positive feedback. After a mere 5 runs I focussed on making no mistakes again, and I was below 25 seconds. Mind you, before that, trying for over 50 runs there was no way in hell that I was going to go under 26, my top times were averaging around 27.5 seconds. By forcing the errors I had broken the pattern and found a more optimum way of behaving. Much faster than waiting for some luck would have done.

What's going on here
When you make a mistake you tend to give yourself negative feedback "aww shoot!" and this tells you to avoid the behavior that led to this the next time. When you don't make mistakes and you have a better time than before you give your self positive feedback and reinforce the behavior that led to the result. When you try something new and it doesn't pay off immediately, you will give yourself negative feedback. We crave positive feedback, so we keep on aiming to get that result improving, even if our method is flawed and we're in the wrong rhythm. It's hard to break a pattern that reasonable results but is suboptimal. The main reason here is that when left to it's own self learning devices, the variation in methods goes down rapidly. It takes a conscious effort or a lucky coincidence to break out of the suboptimum. The lucky coincidence becomes less and less likely the more you reinforce the suboptimal behavior. This is caused by a thing called Extinction Burst which is intended to pulls you back into the behavior that worked in the past.

Download now or preview on posterous
PastedGraphic-3.pdf (21 KB)

What's to learn?
It pays off to try things, even if the first results are not impressive follow through for a bit and see if you might just be crossing a chasm and things are better on the other side. I've seen this pattern not only in my own behavior, but also in the behavior of groups of students and teams of software developers. As long as we didn't plot the graph of all possibilities the only way to find out that behavior is suboptimal is to explore and find a better type of behavior. With Math Workout this is easy, because if I can't make the required time my behavior must be the problem. If there is no required time and nobody knows which goals to set (like in Software development, or any kind of creative work really) you can never be sure of what works best until you've literally tried everything. The trick is to find a way to reward yourself for trying the new thing that doesn't relate too much to the old reward of squeezing every last drop out of a poor method.

I'm not saying you should be trying everything right now, but how about trying one thing each week? That would already be a huge improvement for most of us I'm sure.

Posted via email from Iwein's braindumps

No comments: