Dave Snowden

Liminal Cynefin – image release

RSS Feed

I decided it would be a good idea to make the image of Cynefin available, if only to reduce the change of people drawing it as a two by two and adding a diamond to keep me happy. Typical Saxon tendency, reducing Celtic curves to straight lines. I’ve made it available in Powerpoint as Keynote users can pick that up – it was generated in Keynote. I don’t know why I didn’t do this before but better late than never. I’ll also aim to add more images later and all of them for black and white backgrounds.

I put some effort here into positioning the liminal domain, but I’ll amplify that in a subsequent post where I add in the dynamics along with bullet points for all the shared domain slides. One point I didn’t make in my last post on the subject was the advantage the liminal model gives to those people who find it difficult to grasp the transient nature of the Chaotic and Disordered domains, even those who don’t see a difference. With the liminal version we now have:

  1. A main chaotic domain which is a crisis, the absence of effective constraint and a liminal area which is where managed innovation and sense-making can place.
  2. A liminal area of the complex domain which allows me to distinguish between iterative approaches (such as Scrum) and parallel safe to fail experiments and emergent catalysts in the complex domain.
  3. A final resolution of the authentic and inauthenticaspects of disorder.

So hopefully this helps and I’ll wait to see if people pick up and use the images. If its useful I will add more over time.

The top image by the way its a cynefin one, in these sense of being a place of my multiple belongings. Its the western ridge from Cadair Idris and yes its me, one of my walks with companions. The next time I’m there will be on the 1st March 2018 on what should be day 65 of my round and back through Wales walk. The final three days are mountain days and the date chosen is significant so I hope that Dylan Thomas’s designation of the mountain will not apply that day.

  • Asher Sterkin

    Great images and a really important milestone for Cynefin! The point about straight vs curve lines is also very important. The only one point I would like to challenge is, in my view, too easy dismiss of potential viability of transition between Complicated and Obvious. My own experience with software development and automation suggests that exactly there many interesting things take place. Following the Cynefin core domains classification I would dare to argue that ALL running production software at the moment fails squarely into the Obvious domain. The reason is simple – with all recent hype around AI all what computers could do is to sense-classify-respond. Even with the most sophisticated neural networks software still runs as a closed system, it lacks judgement, experience and professional intuition – all these would be required for performing analysis characteristic for Complicated Domain. Computers could calculate according to selected formula or even to choose one from a predefined set but they could not, yet, choose a suitable formalism for a new problem and this is what experts are, at least in principle, capable of doing. So whenever we want to automate something we basically want to push it into the Obvious Domain. But pure automation seldom exists, neither it should be. In many cases automated processes are augmented with some form of human analysis and decision making (aka Complicated Domain).
    When experts find some recurring patterns they may try to automate its handling by applying Computer (algorithms) or Data (Neural Networks) Science. In many cases human experts still periodically validate the correctness of computer models and constantly tweak and adjust them (this is for example how most recommendation systems work today). Therefore process automation constantly fluctuates between Complicated and Obvious Domains. On the other hand, software development process itself in terms of product/market fit, team structure etc. fluctuates between Complex and Complicated as you have already stated on multiple occasions. These two transitions areas co-exist in parallel which brings up richer and more interesting picture.

    This is how things look for me. If I somehow completely misinterpreted the Cynefin domain definitions I would be happy to know in what.

    Best regards,

    • Dave Snowden

      I’m not devaluing the transition, but I a saying its not a liminal one. I think you are right for a lot of computing, but AI is now capable of performing in the complex domain through self-learning for example.

      • Asher Sterkin

        So my misunderstanding was that of “true liminality” as “state of tension during transition” when things start moving towards emerging direction but options are still open enough to reverse decisions at low cost (hope this back of napkin recap is not wrong). And then you say that “The boundary between the two ordered domains equally is an artificial one based on human perception, agreement and understanding. So it can be crossed at will, but in general the shift from Complicated to Obvious is one to only do if you are absolutely sure about it.” Leaving the self-learning AI aside for a while and assuming that typical business process automation (including DevOps) software is operating in Obvious domain I still have a vague feeling that introducing some dose of liminality would be at least desirable. Too often manual processes being conducted by experienced humans (aka Complicated) are prematurely automated in the worst possible way leading to serious service quality degradation and absurd results of kind “you are right, but the computer does not allow us to do this”. So what you wrote in the firt post on liminal Cynefin still seems to be applicable: “The same applies in reverse moving out of complicated into obvious. That shift is a major commitment to robustness and an assumption that failure is implausible, you can afford the energy expenditure required for rigid order and predictability. So you hold for a long period of time before you make the final abrupt transition. If you go too soon then the switch to chaotic failure, induced by complacency and an inability to detect the weak signals of legitimate deviancy results in catastrophic failure.” When Eric Evans is talking about “tackling complexity in the heart of software” of Domain-Driven Design it’s also about careful and responsible choice of what should be automated in the first place plus gradual introduction and readiness to reverse the decision should it proved to be wrong. Unfortunately the current obsession with mega-scale and exponential growth is a huge obstacle to such attitude.
        Nevertheless, I will respect your decision not to treat Complicated to Obvious transition as a liminal one. For me Liminal Cynefin is unsurpassed candidate approach for dealing with complexities of software industry and big thanks for sharing it with us.

        • Dave Snowden

          Automation of a human process I would always place in the liminal zone from complex to complicated. A shift to Obvious comes when everything is worked out

          • Asher Sterkin

            In many cases I found it useful to keep in mind two transitions simultaneously (kindish of Control/Data Plane separation in networks). One Complex -> Complicated for business model, product features, market fit, org structure and development process. Another Complicated -> Obvious for actual automation implementation and tools choice. I yet to see anything even Complicated in the mainstream enterprise software – it might be challenging, hard to implement, sophisticated but still Obvious from the “sense-classify-respond” dynamics point of view. For example, introducing SCRUM into organization vs using (or not) SCRUM automation tools like VersionOne or Jira. Ditto with regard to test automation and CI/CD frameworks. Would it still be kosher Cynefin?

          • Dave Snowden

            I’d tend to see aspects of ERP systems in both complicated and complex; the ERP system as a whole is too big a unit to consider. Some of the negative impacts of over constraint could be complex or impending chaos. Scrum I generally place in the liminal zone of complex into complicated. Automation, open question really depends a bit on context and human interaction. Better qualified people than me to make that call

  • Russ Field

    With the exception of subdividing Disorder, this line of thinking seems to largely echo the earlier concepts of sub-domains at the borders – in this case, specifically “Complex-going-Complicated” and “Chaotic-going-Complex”. I get the advantage of more formal dissection of Complex-Disorder-Chaotic, but – at the risk of seeming obtuse – can you please help me understand the advantage over talking about sub-domains?

    • Dave Snowden

      mainly the napkin test – one extra line is easier than four three by three matrices

  • http://agilepainrelief.com/notesfromatooluser/ Mark Levison

    Dave since you made this available in a powerpoint I assume its ok to use it in a presentation/workbook for Leadership at a client. The question becomes, what attribution/copyright would you like?

    (It wouldn’t hurt if you added the resulting answer to the powerpoint then people don’t have to guess).