Dave Snowden

Boundary conditions

RSS Feed

For various complex reasons which started with a casual throw away remark at Lean Agile Scotland followed by a brief twitter storm I am spending two days in Dubai with a Kanban group. That includes the co-authors of Essential Kanban Condensed, namely David Anderson and Andy Carmichael. I have the latter to thank for this picture he drew during one of the sessions where I was talking about Cynefin and placing emphasis the boundary conditions, including seeing the complex-complicated boundary as a domain in its own right. To make that easier I used the plane model of Cynefin which was originally draw with the incomparable Max Boisot at the Academy of Management in Washington many years ago. It shows the boundary between Obvious and Chaos as a cliff, in the sense of catastrophe theory.

The reason for doing this is the creation of a deeper understanding of the boundary conditions and their nature. Now I’m not trying to over complicate Cynefin. The basic representation remains as is, but the value of Cynefin is that it can be easily understood in that form, but can also be progressively understood at deeper and deeper levels; this is one of those. So lets look at the four main boundaries, ignoring for the moment the ones to disorder:

  1. Obvious to Chaos
    This is the complacency boundary, the imposition of inappropriate best practice or standard process regardless of consequence. Both wilful and accidental incompetence come into play here. It’s also the failure because of your competence in an old way of looking at things, to reference Clayton Christian’s work. To be very good at doing things the old way can be deeply problematic with the eco-system has shifted. I’ll pick that up tomorrow I expect as I will be talking about dominant predator theory then.  Recovery (Chaos to Obvious) is both expensive and inappropriate so I have described this as a one way boundary.
  2. Complicated-Obvious
    An easy boundary to cross and I portrayed it as a slippery slope.  In the Complicated Domain there is a degree of permitted variation, the system has adaptive capacity with governing rather than rigid constraints.  It is all too easy, especially in consultancy drive improvement exercises to shift from a degree of variety in good practice to a monotone rididity of best practice.  Done well, operating theatres for example, the shift has value but it is a heavy investment to achieve results and there is a lack of flexibility in change in consequence.  So it is a journey that should have a wall with gate keepers, but lack of definition means that it is all to easy to dive into it.   You can make the return journey, but if that is likely it is better the avoid the need and live with some degree of variation in the short to medium term.
  3. Complex-Chaotic
    Again a two way boundary, but each routing is different.  A chaotic system naturally becomes complex as constraints emerge, in fact maintaining a chaotic state (no constraints, no connections in Cynefin terms) takes a lot of energy and resolution.  The reverse is more difficult as systems don’t like randomness.  So creating chaos for the use of a tool like MassSense requires a commitment and as such the boundary needs to be a gate.  If you want a story metaphor think about the town of Wall in Neil Gaiman’s Stardust, named after the stone barrier that separates the town from a grassy meadow (if you haven’t read it do, if you haven’t listed to the Radio 4 adaption do, if you haven’t seen the film with Michelle Pfeiffer and Robert De Niro in magnificent form go seek it out but prepare to be disappointed by Ricky Gervais).  The entry to the world of Faerie is fraught with danger, and for the star to pass is to loose life.
  4. Complex-Complicated
    To my mind the most interesting and one I am thinking of making a domain in its own right in come representations. I’m playing with different metaphors here, but I’m thinking of an interesting valley where it is easy to linger but from which you have to move.  I have one in mind a short walk north of me on the Marborough Downs with neolithic field systems and the Grey Wethers of free lying sarsen stone on the walking path to Avebury.   The transition from complex to complicated can be achieved with the imposition of constraints produces repeatable results.  So parallel safe-to-fail experiments run and create sustainable practice which is taken through linear iteration to refine, correct and scale it (for example SCRUM in Agile).  The shift to complicated is desirable as it allows scale at reduced cost with increased certainty of outcome.  The mistake people make is to stay in the domain when the constraints are starting to fail to produce repeatable results – then we need to move back to the complex.

So at a high level that describes the boundaries. However I am starting to think that the boundary between complex and complicated should at times be maintained as a semi-permanent space – something I have talked about as a grazing dynamic. We should maintain a space of partial certainties, while maintaining options. But that is for a future post.

  • Mika Latokartano

    I was initially skeptical about considering the boundary between complex and complicated as a semi-permanent space or even a domain in itself, but as I thought about it, it reminded me of Dual-Phase Evolution in complex adaptive systems. I don’t consider the boundary to be a domain in itself but the connectivity and phase-space transitions are recurrent so there’s coherence to the hypothesis. Complex-complicated transitions share many similarities with DPE and it might not only be interesting to study complex-complicated transitions as a form of DPE, but provide a framework for further consideration and development of the proposed idea.

    As it would be redundant for me to summarise Dual-Phase Evolution here, I’m linking to the Paperin, Green, and Sadedin paper from the Journal of The Royal Society.

    Dual-phase evolution complex adaptive systems
    Greg Paperin, David G. Green and Suzanne Sadedin

    • tonyjoyce

      This may be pertinent and a way to dive a bit deeper, please see “Cascade Control” discussion in
      Physics Today > Volume 70, Issue 1 > 10.1063/PT.3.3426
      JANUARY 2017 • page 32 PDF FULL TEXT
      “The unfolding and control of network cascades”

  • Dimitar Bakardzhiev

    How does “we bounce off boundaries” from say http://cognitive-edge.com/blog/abide-ii/ relates to the idea implied here that boundaries are to be crossed?

    • Dave Snowden

      If the boundary is there you bounce off it, if the boundary owner opens a gate or makes it permeable, or you break it ….

  • Marcus Guest

    Amplifying the importance of boundaries better demonstrates the depth and utility of the framework, so quite excited by your early thoughts here.
    The Complex to Complicated shift is one where some clients have had the most trouble post-sensemaking workshop: ‘we have probes, now what next?’. Maybe indicative of the (developing) markets I’m working in (where Agile is only just starting to gain traction). But brings to mind vectors on heat maps. Parts of the terrain with no adjacent possibilities suggests movement through these parts will require extreme efforts – like traversing a desert or crossing a mountain.
    The shift from Complex to Complicated seems similar (at least for some organisations). So rather than a wander in the valley it evokes a sense of an ‘overcoming’ mission, where you can’t quite see the other side from the current complex domain, which signals a reason to evolve from it eventually.

    • Dave Snowden

      The lack of adjacent possibles is one of the reasons I was playing with making it a pseudo or temporary domain. I’ll pick up on that in a following post

  • Dimitar Bakardzhiev

    Cynefin has three ontological states (Order, Complex, Chaotic), four epistemological states (Obvious, Complicated, Complex, Chaos), and five phenomenological domains (Obvious, Complicated, Complex, Chaos, Disorder)
    Considering the above which are the boundaries in this post – ontological, epistemological or phenomenological?

    • Mika Latokartano

      All of them, I’d argue.

      There’s a duality between descriptive and ontological complexity (Emmeche, 1997) and in socio-complex systems we are dealing with both. Defining the ontological boundary of a complex adaptive system is difficult because they are open systems, in constant interaction with their environment. How do we determine which parts belong to one system and which ones to another? The boundary isn’t clear and we can define the system arbitrarily.

      Cilliers (2001) wrote “Boundaries are simultaneously a function of the activity of the system itself, and a product of the strategy of description involved. In other words, we frame the system by describing it in a certain way (for a certain reason), but we are constrained in where the frame can be drawn. The boundary of the system is therefore neither purely a function of our description, nor is it a purely natural thing.”

      The boundary, therefore, is phenomenological as well as ontic-epistemic. It cannot be clearly defined, it’s fuzzy and has descriptive and ontological properties.

      SenseMaker can give us a good idea of a system and its boundaries by identifying patterns, attractors in the system, stability and fitness landscapes, and in a continued research showing their dynamics over time. In a broad sense the boundaries have been constrained by the signification framework, but the narrative fragments themselves aren’t rigidly constrained by that either. They can still contain anything. It does, however, provide a way for comparative ontology by combining quantitative and qualitative research, and I think it’s about the best we can do.

      Emmeche (1997) said “we cannot a priori decide whether descriptive complexity entails ontological complexity.” It may, but it may not. Boundaries, likewise, can be descriptive or ontological, or both.

      • Dimitar Bakardzhiev

        Boundaries between Order, Complexity and Chaos ontological states in Cynefin are phase shifts. Your response addresses only boundaries of complex systems. Complex system is only one of the Cynefin three ontologies.

        • Mika Latokartano

          That’s true, I was focusing on arguably the most interesting dynamic between complex and complicated, and should have addressed the other domains as well, but I think the argument is valid for boundaries between other domains as well.

          • Dimitar Bakardzhiev

            That is what I am not comfortable with – the definition of a boundary is intertwined with the definition of the act of crossing it. Boundaries are used interchangeably with the act of crossing them.
            If a boundary is a phase shift between two ontological states then the act of crossing the boundary does not exist because there is no crossing! We actually change the state, no?

          • Mika Latokartano

            Yes, but I suggest we could look at this again as a form of descriptiveontological dualism; the descriptive aspect being crossing a domain boundary, while the ontological aspect being an ontological change, a transition in the state space, a phase shift in the system dynamics. Even highly constrained systems like the Obvious domain don’t exist in isolation and have varying degrees of freedom and connectivity – only constrained – and may be perturbed to induce a transition in the state space, so we don’t even need to limit ourselves to considering only the Complex domain.

  • Pingback: On Ontological and Descriptive Complexity | imaginarytime()