Monday, April 21, 2014

Research and Failure

By Pat Muoio

Folk wisdom has it that a good research program should fail at least 70% of the time. This might lead one to think that research is the perfect endeavor for the lazy and the inept. Yet research remains a respected pursuit; and researchers are generally thought to be driven and accomplished (insert image or your favorite inventor or mad scientist here). So how do we reconcile this drive for truth and innovation with the complacent acceptance of a high failure rate?

First we have to recognize that not all failures are created equal. There is one species of failure that results from lack of critical thinking, misunderstanding of the problem, unchallenged assumptions, poor experimental design, or general incompetence. This kind of failure is no more acceptable in research than it is in development or operations. The desirable species of failure comes from taking significant technical risk and pushing the boundaries of what is currently known. The thinking is, if you go out on a technical limb, it will fail to bear your weight a good percentage of the time. You can increase your chances of being supported by staying close to the trunk, or by only venturing out on the thick limbs that have been around for a while, but you can’t reach very far from these vantage points. To expand the scope of your grasp, you need to explore the less mature parts of the tree.

But falling out of a tree hurts (to torture this analogy just a little bit more) so why is climbing trees a good thing? For one, the view from the top when you are successful is spectacular. For two, you learn a lot about the problem, and about the limits of our understanding, every time you fall. And, if you are self-critical about your climb, your analysis of what went wrong improves your chances of succeeding the next time. This learning, born of risk-taking, is the value of failure in research.

Yet taking risks is not the same as being foolhardy, and it is critical to keep this in mind when embarking on a research activity – good research needs a strategy. You can assess the resilience of the branches of the tree you want to climb. You can tell in advance that some branches are just too weak, or are pointing downward and so won’t improve your view in any case. You can trace a path through the tree that enables you to jump to a nearby branch when you hear the one you are on starting to crack. And you can put a knowledge-collecting net near the base of the tree so you can bounce back up after the fall.

No comments:

Post a Comment