It started, as these things so often do, with a Tweet on Twitter.
(This has turned out to be an enormously long post – I’d better put a ‘Read more…’ link in here before continuing.)
- snowded: NLP as Cargo-Cult psychology. Great paper http://jarhe.research.glam.ac.uk/media/files/documents/2009-07-17/JARHE_V1.2_Jul09_Web_pp57-63.pdf
The link points to a 7-page academic paper [PDF] by Gareth Roderique-Davies of University of Glamorgan, which purports to indicate that NLP (‘Neuro-Linguistic Programming‘ – a kind of self-hypnosis psychological tool) has no scientific basis, and is therefore ‘cargo-cult psychology’. I do take his point that there are some worrying flaws in NLP itself, and even more worrying flaws in many of the ways in which NLP is promoted and used these days. But I’ve seen this kind of ‘scientific’ review before, and I said so in my re-Tweet of Dave’s first message:
- tetradian: @snowded: “NLP as Cargo-Cult psychology. Great paper http://tr.im/IjbF ” <disagree: NLP has serious flaws but this is just a hatchet-job
The problem is that the reviewer is trying to assess NLP in conventional scientific terms – which makes no sense right from the start, though his world-frame would itself make it impossible to see why it makes no sense. (For enterprise-architects, by the way, this is the same underlying reason why IT-centrism or organisation-centrism is such a problem: the frame itself makes it impossible to see beyond the frame.) The title of Bandler and Grinder’s original book that defined NLP way back in the 1970s gives the reason why the scientific frame won’t work: it’s called The Structure of Magic.
Yup, that’s right: magic.
Most self-styled ‘scientists’ treat that word in the same way as IT-centric ‘enterprise’-architects treat business-architecture and beyond: namely a randomised, undifferentiated grab-bag of all the bits of reality (or business-reality, in IT-EAs’ case) that they don’t understand. And then complain that it’s a mess, and doesn’t make sense in their own chosen terms, and therefore doesn’t exist. Which is not exactly honest – and it’s certainly not helpful in practice, because magical-thinking is often the only way out of many everyday scientific, technological and business dilemmas and problems.
A small tale here. Everyone ‘knows’ that Isaac Newton was one of the world’s greatest scientists, yes? (Which he was, of course.) But not many people know that he was also interested in a great many other subjects, including religion, alchemy, astrology and much else besides: in fact he wrote more on alchemy, for example, than on all of his scientific studies put together. Edmond Halley, the then Astronomer Royal, was berating Newton for the latter’s studies of astrology: it was all nonsense, he said, ridiculous, utterly unscientific – or words to that effect, anyway. Newton’s short, sharp retort: “I have studied the subject, sir, and you have not!” End of conversation…
Which brings us back to NLP, and the structure of magic. As it happens, I have indeed “studied the subject, sir” – for more than forty years, in fact – and I guess most people reading this blog probably haven’t, so it might be useful if we do a quick tutorial here on the role and limitations of the scientific frame and mindset, and the contrasting role of magical-thinking. To do this I’ll pick up on another of today’s Tweets, from knowledge-management (KM) guru David Gurteen:
- DavidGurteen: Is KM a Pseudoscience? #KM http://www.greenchameleon.com/gc/blog_detail/is_km_a_pseudoscience/
This link points to an article by another key figure in KM, Patrick Lambe – much better thought-through and much more considered than the previous piece. Using a checklist from an article by Barry Beyerstein, he scores KM overall as having a score of only 5.4 out of 10 as a ‘rational endeavour’, and concludes that it is too close to a pseudoscience: “must do better”, he says. But what that article misses, yet again, is the bald fact that trying to assess most of KM in scientific terms makes no sense. The only way we can make sense of it is via a magical approach.
(Yes, I know I still haven’t explained yet what I mean by “a magical approach” – give me a chance, I’m getting to that in a moment! )
Before we can look at magic, we need to understand science – as much for what it isn’t as for what it is. What it isn’t – as any competent scientist would admit – is “the answer to Life, The Universe, Everything”. Instead, it’s a particular body of knowledge, developed in terms of a specific set of methods and assumptions, and which can only make sense – or be useful and valid, rather – within a very specific set of constraints. Science has been extremely successful within those constraints - so successful, in fact, that many people fail to realise that by its own definitions it is not and cannot be successful outside of them. Therein lie many huge problems for KM, for enterprise-architecture and for many other disciplines – including magic.
This is perhaps best described in one of my all-time-favourite books, WIB Beveridge’s The Art of Scientific Investigation. First published in 1950, it’s been continually in print ever since, and remains one of the great classics of scientific research. I’ll have to quote from memory, as my copy is back in Australia, but his introduction starts like this:
Complex equipment plays a central role in the science of today, but it should never be forgotten that the most important instrument in research must always be the mind of the researcher.
Beveridge expresses concern that “perhaps not enough attention is paid to making the best use of it”. To this end he focusses on the actual practice of science, rather than solely on the end-products of that practice. Hence his book includes detailed descriptions and examples on strategy, hypothesis, the use of chance and intuition, and “the hazards and limitations of reason”. (Most of his examples come from his own field of biology and biochemistry, but they’re just as applicable to every other branch of science.) The summary in his chapter on reason is particularly important, though forgive me if I again have to quote from memory:
The origin of discoveries is beyond the reach of reason. The role of reason in science is to come afterwards, to review and reassess and to build a general theoretical scheme. … Most biological ‘facts’ are so uncertain that at best we can only reason on probabilities and possibilities.
And that last sentence remains just as true as ever, despite the advances of molecular biology and the like over the past half-century: the only certainty in science is that many things will always remain uncertain. But it’s all too easy to forget that fact: that’s where the problem starts.
It’s also all too easy to forget that ‘the scientific method’ depends entirely on its base-assumptions: it cannot be relied upon outside of their remit. For our purposes, the most important of these assumptions include:
- causality – all events are connected via cause/effect chains in a linear ‘arrow of time’
- repeatability – given the same conditions, all experiments and results must be repeatable by others
- falsifiability – every hypothesis must be framed in such a way as to enable its negation by experiment
- consistency – the results and hypotheses in each domain of science cannot contradict those of other domains of science
Within those constraints, science works extremely well – and likewise, usually, any technology based on that science. But it’s essential to realise that it only works within those constraints – and there are plenty of conditions where those assumptions break down. Repeatability and falsifiability will seem to make sense whilst we’re dealing with the mid-range of scales, but in fact they break down as we move more towards the very small – down into quantum levels, as per Heisenberg’s Principle - or to the very large – where experimentation and repeatability are often inherently impossible (at least on the kind of time-scales that we live in!). The same applies as we move more towards unique events: chaos-mathematics makes the level of unpredictability more predictable, but does not reduce the unpredictability itself. Consistency also frequently breaks down between domains: last I heard, for example, the most likely theory of star-formation requires a universe much older than ‘permitted’ by the most likely theory of cosmology. And out at the fringes of science – particularly in nuclear physics – there are plenty of examples where any linear concept of causality will break down, and at times looks remarkably like traditional magic. For example, the old magical notions of ‘action at a distance‘, teleportation and telepathy are all ‘permissible’ in current quantum-entanglement physics, and in some cases have even been demonstrated in laboratory-experiment – even if only at quantum scales.
And there are plenty of real-world, everyday examples of where those assumptions will break down – especially in KM and the like, where we’re often dealing with contexts which, by definition, are either unique or near-unique. So complaining that KM might be considered by some to be a ‘pseudo-science’ is to miss the point, because there’s no way that it can be a ‘science’ in those formal terms above. Instead, to make sense of what’s going on, we may well need to turn to other approaches: science is one approach that we might use, but it’s not the only one.
Which, by a round-about route, brings us back to where we started, with Dave Snowden and the Cynefin framework. Starting from the unknown – what Dave describes as the domain of ‘Disorder’ – we have four distinct methods to ‘make sense’ of what’s going on:
- the Simple domain: apply rules to ‘categorise – sense – respond’
- the Complicated domain: apply algorithms and logic to ‘sense – analyse – respond’
- the Complex domain: apply guidelines and heuristics to ‘probe – sense – respond’
- the Chaotic domain: force change through action, to ‘act – sense – respond’
I know Dave can get ‘curmudgeonly‘ when we place these Cynefin domains in a simple two-axis frame, but in this case there’s one frame that aligns extremely well, and does add quite a lot to our understanding of Cynefin itself. These two axes are value versus truth, and inner (personal) versus outer (collective), which gives us four domains: inner truth, outer truth, outer value, inner value. These domains map almost exactly to those four main Cynefin domains and their sense-making tactics:
- ‘inner truth’: Simple domain – rules or supposed ‘universal truths’ that purport to apply to everyone, everything, everywhere
- ‘outer truth’: Complicated domain – algorithms and the like, often with multiple factors and complicated interactions and delays, but always amenable to causal analysis
- ‘outer value’: Complex domain – use ‘seeds’ and experiments to probe into the context, to allow meaning to emerge
- ‘inner value’: Chaotic domain – any meaning that may be derived is context-dependent and probably personal only
(The chapter ‘Can’t we explain this scientifically?‘ in my 1990 book “Inventing Reality” likens each of these modes with a means to survive within a swamp: run too fast to sink; climb up a pole; weave a platform between a group of poles; or spread your weight on swamp-shoes. The advantages and disadvantages of each mode are summarised in some detail there: might be worthwhile to read that chapter now and then come back here.)
In practice we would – or should – usually switch between each of these modes, much as Beveridge implies in The Art of Scientific Investigation. But the key point here is that a ‘scientific’ approach – which depends on causality and logic – can only make sense in the two ‘truth’ domains. Trying to use ‘truth’ tactics in the ‘value’ domains is not a good move: we risk ending up with what Dave Snowden calls ‘pattern entrainment’, such that in effect we use a quasi-religious belief as a substitute for true science or sense – which is not a good idea. (For more on this, see, for example, Amory Lovins’ video on “How the practice and instruction of engineering must change“.). Which means that we need to use entirely different approaches in the two ‘value’ domains. We could use terms such as ‘non-rational’, ‘arational’ or ‘meta-rational’ for this, but we might as well use the term that already exists for this: magical.
Magical-thinking isn’t a mistake: it’s what we need to use in the two ‘value’-domains – or, in Cynefin terms, the Chaotic domain and, especially, the Complex domain.
This post has rambled long enough already, so I’d better not go into too much detail. But one of the key tactics here is to deliberately use beliefs as tools, especially in the Complex domain, using them as if they are true whilst still recognising that they may not necessarily be ‘true’ in absolute sense. In classic scientific terms, another name for this tactic is hypothesis, as contrasted with idea (Chaotic domain), theory (Complicated domain) and law (Simple domain). It’s what we do in most technology-development: for example, we might use ideas from science, but we might also use analogy, metaphor, serendipity or even images from a tarot-deck – what works is whatever happens to work. And the fundamental question here is not science’s ‘How does it work?’, but ‘How can it be worked?’ – not how do we make it more ‘true’, but how do we make it more effective, more efficient, reliable, elegant, appropriate, integrated.
(Incidentally, this is one of several reasons why using the term ‘applied science’ as a synonym for ‘technology’ is misleading and even dangerous, because we end up applying the wrong criteria to measure that technology’s value – assuming ‘technology’ in the original sense of ‘tekne‘, a body of knowledge and related practices rather the rather incomplete sense as ‘machines and stuff’. Another concern is that by purporting to be ‘science’, a usage of technology can also attempt to claim science’s status as ‘value-free’ – and hence supposedly not subject to the ethical and other value-constraints that, by definition, are actually the core of every technology. And magic too, for that matter . In this sense, technology and science are fundamentally different from each other, whereas technology and magic are fundamentally the same. In fact the only real difference between the latter is that magicians tend to be a bit more ‘way out’ in their choice of beliefs, especially when the technology is more about mind than matter.)
Whichever mode we use at any given time, the key to all of this is discipline. (This applies in magic as much as in any other technology: as the pseudonymous author of the influential SSOTBME put it, “all those boring meditation books are just the magical equivalent of a school chemistry primer”. But that’s another story… ) Which, finally brings us to why I wrote this post in the first place, because we need a disciplined approach not only to the use of each domain, but also to how not work work within each domain, and how instead to switch between the domains in an effective, intentional manner.
Most readers of this blog would know me as a specialist in whole-of-enterprise architecture. But my real interest, and real work, is in methodology and meta-methodology – the design of methodologies to suit each specific context and need. Behind that, what really concerns me is the process of developing skills as true skills capable of dealing with the complexities and chaos of the real world – rather than as glorified ‘trainings’ that are only usable in the safe, easy purported-predictability of the ‘truth’ domains. I’ve been engaged in this work for well over forty years: for example, one of the tools I developed that you may have seen is the Skills Labyrinth, a live-metaphor for the skills-learning process.
But one of my primary test-cases for this – mainly because it’s almost the closest I can find to a ‘pure’ interpretive-skill, with very little manual-skill and technical-knowledge required to get started – is what’s known in Britain as dowsing, the generic for ‘water-divining’ and the like. (Each country has their own term for this: Americans would know this as ‘water-witching’, for example, whilst Dutch might call it ‘wichelen’.) It’s a classic ‘magical’ skill, sufering – as so many do – from an overdose of idiots, and much-derided by self-styled ‘skeptics’ who rely only on ‘scientific’ theory rather than technological practice and hence don’t have any real grasp of what they so obsessively dismiss. (As it happens, we know a great deal about the physics, physiology and psychology of the skill: one key point we now know for certain is that there is no single mechanism involved, but rather a complex ‘weighted-sum’ merge of multiple mechanisms. Hence most of the classic means of scientific enquiry – “how does it work?” – make little sense, whereas technological enquiry – “how can it be worked – does indeed work well here.)
Worldwide, I’m actually better known as a writer on dowsing and related subjects than on IT or enterprise-architecture: my first book on this – nowadays known as The Diviner’s Handbook – was first published in 1976, translated into some dozen languages, has been in print continuously ever since, and is regarded as one of the standard reference-works on the subject (or learning-guide, rather, because that’s its real purpose). And I apply exactly the same rigour to my work in that field as I do to anything else: I insist on keeping myself, and others, strictly to the correct discipline in the appropriate domain. Which at times is not ‘scientific’, of course – but so what? If the ‘scientific’ mode is not appropriate in that part of the technology, don’t use it! Which is exactly the same principle as we need to apply in KM, or enterprise-architecture, or anything else that is inherently complex and in any way inherently unique, and hence where the usual constraints of ‘rational repeatability’ and the like do not and cannot always apply.
Hence yet another book of mine, co-authored with the archaeographer Liz Poraj-Wilczynska, and published late last year, called Disciplines of Dowsing. (You can download the e-book version for free from the website, though please consider buying the print version if you’re going to use it in practice!) Parts of this work have also been published in the Berg peer-reviewed academic journal on archaeology, Time & Mind. In it we explore the application to dowsing practice of the same four approaches to sense-making and action, linked to Cynefin as above, and cross-linked to standard quality-improvement tactics such as kaizen, the Deming/Shewhart PDCA cycle, ISO-9000:2000 and reflective methods such as After Action Review. It’s the same principles, applied in a slightly different area to what most KMs and EAs might know, but otherwise no different at all. What is different – and which we haven’t seen anywhere else – is an explicit emphasis on how and when and why to switch between each of the disciplines. Which, in turn, we could – and, I would argue, we should – apply in turn to our other everyday work-domains such as KM and EA and the like.
There’s also a strong emphasis in the book on how to identify and avoid some all-too-common pitfalls, the ‘seven sins of dubious discipline’ such as the Hype Hubris, the Newage Nuisance and the Meaning Mistake. (‘Newage’ is perhaps a more accurate term for much of what purports to be ‘new age’: it rhymes with ‘sewage’, ‘the discarded remnant of what was once nutritious’… yup, I can be a cynic too! ). But the point here is that, again, there are exact equivalent of the ‘seven sins’ in every other kind of skill, including those in the sciences: for example, Roderique-Davies’ paper on NLP includes several all-too-obvious examples of the Meaning Mistake. If we don’t understand the limitations of science, and worry too much about seeming ‘unscientific’ or ‘pseudoscience’, we’re likely to end up damaging the quality of our skill and our results rather than improving it. In that specific sense at least, magic is real – and as Cynefin shows us, it matters just as much as science and the like to the quality and validity of our practice.
In addition to the e-book of Disciplines of Dowsing, there’s also a two-page reference-sheet that summarises the four sets of disciplines, and that’s perhaps more immediately usable in practice. (The material on the ‘seven sins’ is only in the book, though.) It’s written for dowsers, of course, but it doesn’t take much translation to apply it direct to KM, EA, software development or any other complex-domain skill. Download it, perhaps, and let me know how it works for you? And thence it might be worthwhile writing another version specifically for KM or whatever. Something different to play with, anyway.