One of the key aspects of managing a complex system is the switch from fail-safe design to safe-fail experimentation. The nature of a complex system means that we cannot know in advance what will happen, so we cant design for a desired outcome. Instead parallel and possible contradictory experiments allow us to test the evolutionary possibilities of the system. If an experiment starts to produce beneficial results then we amplify it, if negative (and here safe-fail comes in) we dampen it, or even attempt its destruction. Now most people, other than obscurantist post-Popperniarian positivists (you know you are) get this. But the question then follows, how do you select the experiments? What are the criteria of justification under conditions of uncertainty?
My normal answer is two provide two criteria for an experiment to be run.
- Firstly we can apply a test of coherence. Is the idea for the experiment coherent with a perspective on the systems past and future possibilities? To quote Thagard: A belief is justified not because it is indubitable or is derived from some other indubitable beliefs, but because it coheres with other beliefs that jointly support each other. Note that in validating an experiment I do not have to prove that it will work, but I have to show that it is consistent with a view of what has happened and what could happen in the future.
- Secondly, if the experiment through failure or success would supply useful knowledge about the system at an acceptable cost then it is worth running. Of course I must be able to reverse quickly (safe-fail again) and I must be able to monitor the impact as the experiment runs. But otherwise knowledge is a valid or coherence goal of itself.
Now in both these cases we need to be aware that we will be running parallel experiments in the expectation of substantial failure. That is however more desirable than a one time bet, in which we attempt to forecast outcomes where the future is unknowable, or where we suffer analysis paralysis not able to make any decision without guarantees of success.
Neither is it about gut feel and inspirational leadership. It is a structured approach based on low risk discovery. There is a science to managing uncertainty and theories of coherence are central to the develop of that science and its practice.
Comments (1)
I watched Gordon Brown (PM) on TV this morning say that "there was no national policy for rubbish collection taxes across the UK - only local councils who, if they so desired, could try out small scale initiatives which if successful might then be delivered as government policy. But it will probably take a long time"
Now I don't believe he has too much understanding of why this is a good approach and he emphasised the 'long time' to put off his would be detractors, but it struck me immediately as consistent with your safe-fail approach.
Posted by Ron Donaldson | May 15, 2008 11:19 PM
Posted on May 15, 2008 23:19