Dave Snowden

The necessity of partial constraint

RSS Feed

iStock_constraintBottle000007803521XSmall.jpg At LSSC11 I used the constraint based definition of different system types: Ordered, highly constrained; Chaotic, no constraints; Complex, partial constraints, coevolution of constraint with agent behaviour. In social systems this is generally the easiest way to get people to understand. It also should, but does not always, get people into a position where they realise that the chaotic domain is a temporary or transitionary state with no stability. I’m working on a new framework at the moment which will expand the complex domain into different phases or aspects based on the nature and balance of co-evolutionary forces but that is a few months off being ready for exposure even on a safe-to-fail probe.

In the context of LSSC11 I made the strong point that self-organising teams had to be considered in the context of constraints not anarchy. Spontaneous self organisation having a poor record in the history of humanity and a close resemblance to spontaneous combustion! This was part of a general theme that managing in a complex world requires more discipline that in an ordered one. In the latter we can fall back to recipes and go with the process; in the former we have to think anew, act anew to quote Lincoln.

It is also the case that without some constraints we get no meaning, no cohesion, no evolution. At LSSC11 I used one of my favorite methods, social network stimulation (SNS) as an example of this. In SNS constrains are provided through the rules for team formation and the intractable problems presented. Rewards provide incentive and all three components might be considered affordances (more on that subject tomorrow). In the context of lean software development I suggested the use of SNS to handle the 5% of any project which creates 95% of the problems. In this application of SNS the rules for team formation relate to both technical competence and user understanding, the intractable problems are those aspects of the system which defy precise definition or even the more open definition required in a prototype. An SNS then allows any self-assembled (but within constraints) team to create a solution or part solution to the intractable problems that makes the user community aware of different technological or design possibilities. Those that gain traction are continued those which do not are killed or modified. The reward is the right to continue to work on the successful experiments with a team you have chosen to be a part of.

This approach also satisfies the three heuristics of managing a complex adaptive system:

  • Finely grained objects: small teams, small disposable experimental prototypes/demonstrators
  • Distributed cognition: many different experiments with a divergent range of people involved
  • Disintermediation: senior managers get immediate visibility of successful experiments

Now if we just allowed people to create any team and work on anything they wanted (OK I know its never that extreme but you get the point) then I might not have sufficient variety in the network, or the individual team to get enough of the field scanned. One of the principles of SNS is to get people working with people they are not familiar with as novelty increases attention. Self selection of the team means that negative tension is reduced compared with being allocated to a team. Humans are very good at working out in seconds who they can get on with. Sometimes the chemistry works, sometimes it won’t and that is a matter of human judgement not the primitive witchcraft of psychometric tests.

As constraints in a system increase then the pattern of interaction becomes more predictable, and the degrees of freedom for agency reduce. We shift from exploration to exploitation. In the example above as safe-to-fail experiments succeed increasing investment and focus produces this effect. Over constrain a system however (see illustration) and the only alternatives left are suffocation or catastrophic shattering of the container. Probably one of the most critical things than managers have to learn in this post systems-thinking era is the ability to manage constrains and to do so contextually. More on that in future posts.


PS: Just for the record this posting is not about Goldratt’s Theory of Constraints nor for that matter is it about forms of sexual bondage and the like.