Go uses a 19 by 19 Grid, 181 black stones and 180 white stones. While the average number of legal moves in chess at any point is 37, in Go it’s between 150 and 250 and rarely falls below 50. The most powerful computer around today would require 5 days to calculate all possible combinations of the next 8 moves (5.12 x 10²⁰ combinations). While computers can defeat the best human players, in Go they only manage an intermediate amateur level. The different values of chess pieces make it easier to calculate position, in Go it is far more difficult; the placement of one stone early in the game has an impact on play 100s of moves later. Chess is complicated, Go is complex and the differences give us a way of understanding the different strategies on of robustness and resilience which I raised yesterday.
The basic principle of Robustness is to prevent failure. With resilience we assume that failure is more probable and aim to recover fast. Remember this is not an either/or, its a both/and. Having said that two things are important:
- Focusing on robustness may make us over confident that we can avoid failure and thus reduce our resilience;
- As failure becomes increasingly likely then triggering it early may reduce the negative impact of a later and potentially catastrophic failure;
- If the nature of the system is such that failure is going to happen, moving in early and recovering fast will have competitive advantage.
Resilience does not imply fatalism, we need to do our best to understand the emerging possibilities of the present. Prediction is difficult or impossible to achieve in a complex system but anticipatory awareness should be mandated. So what can we do?
- Shifting from big picture scenarios created by specialists to micro-scenarioes generated by large volumes of players. I posted about this at length some time ago so I will not repeat myself here.
- Using that work to focus on recording and monitoring for outlier events. Gaussian approaches seek to eliminate outliers, pareto distributions take them into account. Outliers are both risk and opportunity.
- If mass participation in creation of micro scenarios is established, then mass participation in the generation of multiple micro-perspectives can be activated during the build up to, and the recovery from a crisis.
- Small units with distributed command and control are both handle novelty better than large hierarchical organisations. Putting crews into place is one way to achieve this. Role based entities not depending on prior knowledge the individual can handle sustained pressure better than hierarchies. They are also more resilient (sic) to loss of personnel.
- Distributed technology and other support to operational units is another obvious change. The BBC, trying to recover back episodes of Doctor Who (they originally overwrote tapes as they were expensive), suddenly discovered that people who have illegally pirated tapes of back episodes became a source of recovery. Over structured and constrained systems are not resilience, allowing some degree of freedom and redundancy increases overall resilience.
- It’s more important to understand that sudden change is likely than it is to predict the nature of that sudden change. It’s also easier. Our use of fitness landscapes discussed in the above referenced post on scenario planning is one way to achieve this.
- The past can inform the future. Gathering large volumes of experiences, both direct and indirect, from previous disasters encountered and avoided by the organisation and by other organisations, all creates a knowledge database that allows the rapid recall of fragmented knowledge from multiple events that we can conceptually blend with our current experiences of failures to plot a way forward. In Iraq it was platoon commanders blogging that worked, not doctrine. A modern system combining micro-scenarios of possible futures and the micro-narratives of past experience gives us a knowledge based strategy for uncertainty that reflects the naturally evolved capabilities of humans.
The above list assumes a degree of prior knowledge of Cognitive Edge methods and approaches, but I wanted to keep it short. I also hope the link to Go is fairly self-evident. If it isn’t then remember that in go all the pieces have the same value, and the context of their position is more important than their nature.