Tests for coherence

As promised yesterday I will return to the issue of constraints and constraint mapping in a series of posts over the next few weeks – hopefully to consolidate them in January. The approach and the method are in development and I used the base structure to good effect in Ottawa today at a course on complexity in Government – something we will next run in Wellington early next year. I’m going to be playing with this for a few weeks before it really stabilises and part of that stabilisation will be some new definitions of the Cynefin domains, and as critically sub-domains. It won’t be fully coherent for some time.

Coherence is a key word in the whole field of anthro-complexity. I was using it for some time when Max Boisot issued one his frequent edicts and told me to read Thagard’s classic on the subject which was helpful. I also had the privilege of being part of a seminar with the author in Paris a few years ago. I don’t fully agree with everything he says, but I strongly recommend the book. For me the value of coherence is that it allows for contingent truth, a determination that while we don’t know everything we are at least going in the right direction. My favourite illustration of this is to say the while we know evolutionary theory is incomplete, and much of what we now know may need radical change, it is coherent to the facts and thus a legitimate direction. Crude creationism on the other hand is incoherent to facts and not worthy of any intellectual effort. Coherence gives us a simple way of getting out of the nonsensical post-fact argument which picks up on an error or change of mind and uses it to justify rejection of the whole, or an all hypotheses are equally valid type of argument.

We use coherence as a criteria for decided what safe-to-fail experiments should be selected for investment in the portfolio approach to action based insight into the emerging nature of a complex system – something which is basic teaching on most Cognitive Edge courses. Increasingly I am being asked what it means to say that something is coherent. So, over the last six months or so I have been developing a series of tests and/or methods for this. The list is not complete, and I have a need to reread Thagard and other material when I get a chance. But for general edification and comment I offer the following and I think (but I am not sure) that the first and second are both necessary and sufficient conditions, while the second is simply sufficient. As I say I am not sure, as an experiment is designed to explore so there may be no necessary conditions.

1. Are the underlying assumptions behind the intervention compatible with what we know from natural science? Is their design consistent with our understanding of the nature of human decision making and system interaction? This may be a negative test as much as a positive one. If the intervention assumes that all respondents will examine our full proposal rationality and make a decision based on their own self interests then it fails. If we assume a measurement system can be come a target it fails, especially if we need intrinsic motivation. This is all at the heart of the naturalising approach to sense-making, which uses science to provide a coherent spine to understanding under conditions of uncertainly.
2. Are there other examples of where this type of approach has created better understanding.  This is a legitimate use of prior cases.  We cannot use our limited understanding of past cases to create deterministic recipes as to future action, but we can use them as a guideline to what may be a legitimate experiment.  Taking contradictory cases (never difficult to find) or assuming a paradoxical interpretation can increase the resilience of any solution but its a good test.
3. Is the proposed intervention consistent as a development phrase from where we are?  This is important and much neglected by those who focus on a desired future state as a starting point.  The gap between the present and that state may simply be too much.  This is not to say that we bound by the past, but that the past provides a constraint on what is possible in the present.  Finding examples of radical shifts in the underlying exemplar narratives of the organisation’s past can be used to enable more speedy or radical changes in the present.  But in general there needs to be continuity, evolution is progressive, occasionally serendipitous but you can’t rely on that happening
4. Does it pass a series of challenges using the ritual dissent technique?  Successive rounds of ritual challenge makes for more resilient experiments, less likely to failure catastrophically, more like to mutate as circumstances challenge initial assumptions.  Increasing the diversity of the challenge and taking a options based approach to the experiment, holding commitment to aspects as long as possible, also helps.  The degree to which there are options to change is a linked test for coherence.
5. Finally we get to MassSense which is a capability within SenseMaker® now fairly readily available, but we do have more work to do on automation.  Here an idea or concept or theory is presented for mass signification and the results presented graphically as a landscape.  That means that dominant views are visible, but also outlier clusters.  A dominant cluster is rather like “ask the audience” in a popular gameshow, it gives comfort but it may not be right.  An outlier cluster is an indication of different thinking.  Both legitimise an experiment.

Now remember I am talking here about coherence in the context of designing exploratory safe-to-fail experiments. That is a special case of a wider application of the idea. I’m also remembering something I was taught in physics and maths – if a solution is aesthetically pleasing then its a good test that you are going in the right direction. Hence my choice of image.

The image for this post is subject to creative commons – my thanks to its originator for making it available on Flickr.