These are my notes from the Cynefin Course I attended at Amsterdam in May 2016
There needs to be a shift away from thinking in terms of robustness to resilience.
These two terms were defined by Snowden as follows.
Robustness: Prevent failure.
Resilience: Early detection and fast recovery.
Traditional risk management is based on the bell curve, also knows as the Gaussian distribution, the red curve below.
We seek to manage risk inside the bell curve, outside of the bell curve are the black swans, the outliers, low probability but high impact events.
These are the events that happen so infrequently that we simply cannot be expected to plan for or manage them, so we don’t.
However in nature events, such as earth quakes, floods or volcanoes follow a power law or Parato distribution, the blue curve below.
If we overlay the two curves we see the following: taken from cognitive edge here
We can see that the Parato curve has a much fatter tail, which indicates that the low probability high impact events are actually medium probability high impact events.
This simply means that these kinds of events happen a lot more often than you would think, indeed much more frequently than we can hope to manage, Snowden referred to this as the New Orleans Levees problem.
So instead of trying to be robust and stop things from happening, we need to acknowledge that instead we need to be resilient, to detect events early and be able to recover quickly.
Research and Monitoring
As we sit in the Gaussian world we have a large number of probable & actual events, therefore a Guassian distribution works well. So we can use a deductive approach to reasoning and say “if we do this it will produce this result with a high probability of success”.
People like to be here!
Double log scale of frequency of event against size. and we see the power law line again.
As we move into the borderline between Gaussian & Parato we switch from the realm of what is possible to what is probable.
We have a range of possible outcomes and its hard to assess them in terms of probability, so here we would use scenario planning such as the delphi method so we can cope with hypothesis and human judgement.
Snowden referred to Clayton Christian at his point.
The further we move to the right and deeper into the world of Parato we go we move into the space of null hypothesis research and abductive reasoning.
The link back here to the Cynefin framework is in the complex space and that in this world we need to be running coherent safe to fail experiments, because we cannot predict the future