On Failure, Success, and Learning
When I was a kid, I decided to invent a new kind of battery. I had a pretty good idea of what was required, having cut open my share of batteries and even built them with a lemon, copper, and zinc. It’s just a matter of two metals (or one metal plus carbon) and a corrosive liquid. How hard could it be to create the battery of the future?
I mentioned my aspirations to my father, who was a chemistry professor. “What do you know about valence?” he asked.
He proceeded to explain about electron clouds and the tendency of atoms to fill or empty their outer ring of electrons.
“So the valence of oxygen is 2.”
“Yes, except when it’s 1 or 4 or 6 or some other value. It’s not always simple.”
I’ve been thinking about that conversation since the end of the Agile 2011 Conference. Kevlin Henney gave a great keynote talk about, among other things, learning about code by reading successful code. “We’re not wired up to learn from failure.”
Yes, that’s absolutely right.
Then Linda Rising gave her keynote (which I unfortunately missed, having to leave for the airport) where where exhorted us to learn from failure. “Try again. Fail again. Fail better! – Samuel Beckett”
Yes, that’s true, too. How can I agree with two seemingly contradictory points?
That’s what made me think about my plans to invent the next-generation battery. Had I continued with those plans, I could have tried 10,000 combinations and been no nearer to success than when I started.
Yet Thomas Edison did invent a successful light bulb. After many failures, he was reported to say, “I have not failed 10,000 times. I have successfully found 10,000 ways that will not work.” If that was all he knew, then he’d have been little better off than I was with my battery. In fact, he’d learned more than he admitted.
During the course of those 10,000 experiments, he built up a model of what was needed and how various materials did and did not fit that model and supply those needs. If we’re paying attention and are very persistent, we can learn from most series of failures as long as each attempt varies from the others.
It’s just not very efficient that way.
We can turn a series of failures into a string of successes quite simply. Instead of expecting each attempt to produce a working light bulb or battery or whatever it is we want, expect each attempt to be an experiment that will teach us something and refine our model. A successful experiment is one that teaches, whether or not it confirms our expectations.
Looking at it that way, we’ll want to design better experiments to maximize the learning rather than trying to minimize the number of trials. This is the heart of “Fail better!”