A Science of Enterprise Architecture? Really?

Filed in Episode 2 by on April 5, 2013 5 Comments

I’ve seen quite a few discussions over the last couple of years about whether enterprise architecture (EA) is amenable to disciplined, rigorous methods – sometimes described as a scientific approach. A lot of what I’ve seen strikes me as anything but scientific. Science and determinism are not the same thing.

What really triggered this blog, though was reading Werner Heisenberg‘s Physics and Philosophy. I can seriously recommend this book for insomniacs but, if one reads it whilst fully awake, it’s a totally fascinating book, which I’ll devote more time to in another blog. The point is that it helped me to understand what it means to apply scientific method. Any attempt to be scientific about EA (or for that matter to reject the very idea) requires us to understand what that really means.

I was born in 1950 (yes, I admit it). That means I grew up in a period when science was cool. It was the road to a new and gleaming future. Scientists were benevolent chaps (yes, they were all chaps) in white coats whose wisdom would free us all. They were hip. Nuclear power was clean and safe and would deliver freedom-giving electricity into every home and office. There’d be no more terrible mine accidents and no one would die early from black lung. In fact no one was going to need to die of anything apart from old age.

hippy

a portrait of the artist as a young man

Then in the late 1960’s and 1970’s science became seriously uncool. Scientists were unhip . They’d given us the bomb and Three Mile Island and didn’t speak out about it: “I only invented the bomb, I didn’t drop it”. There was no new Bertrand Russell. And they had short hair and didn’t listen to our kind of music. I was pretty typical of my generation in that regard.

Recently, though, I’ve come to an accomodation with science and scientists. A fair number of the most interesting scientists turn out to have shared my concerns all along (and played in rock bands). So I’ve been trying to get my head round all the stuff I missed in between times – particularly in physics.

My academic background, however, is in mathematics. It’s the only academic subject I was actually good at. It’s all a long time ago and I’ve forgotten most of what I learned but a couple of things have left their mark on me. One of them is an attraction to rigorous argument and a serious impatience with poorly thought through argument – even when I agree with the conclusions.

So I have a natural inclination towards the scientific approach. The question, however, is what that would mean in an area of practice that’s a lot closer to the social sciences than to, say, physics. A lot of people’s idea of science is attaching numbers to things and, even better, finding a formula that, given any input, will generate the “right” answer. It’s the 42 principle. A while ago someone tweeted “you can’t understand what you can’t measure”, to which I replied “and you can’t measure what you don’t understand”. To some degree both answers are (simultaneously) true.

Until about 100 years ago scientists believed that in the hard sciences we can always (given enough time) arrive at some equation or set of equations that under any set of circumstances will precisely predict the behavior of a system. In other words the behaviour of any physical system must be deterministic. Today that is not always true and besides, enterprise architecture isn’t hard science. An enterprise consists not just of machines with predictable behaviour but also of people. Everything the enterprise does and how it does it is decided by people, all different kinds of people with different interests in the enterprise and different external factors that influence their decisions. So an enterprise is inherently complex, if not chaotic. And it’s not even that simple 🙂 The behaviour of the enterprise and therefore the story and structure of its architecture is affected by micro and macro economics, by policy and regulation, by availability of natural resources etc. So even if we thought economics, politics and sociology could be understood as physical systems, the level of complexity produced by the combination of different factors in play would make deterministic solutions impossible.

Both Tom Graves and Dave Snowden wrote very relevant articles recently about the dangers of trying to control or simplify the inherently complex and about understanding uniqueness and chaos.

Even such an eminent  physicist as Werner Heisenberg didn’t think  “soft” sciences, even biology, could be (adequately) explained by any set of equations in the way that physics can. Getting back to modern physics, quantum mechanics has taught us that elementary particles don’t follow the same rules as everyday physical objects. There are still mathematical equations for then but their behaviour is not deterministic. The theories of special and general relativity have changed our conceptions of space, time and even something as down to earth as gravity. To understand this new world our approach to science has been forced to move away from determinism and the hegemony of the measurable. You can’t measure everything and there’s no point in even trying if you don’t understand what the measurement might mean. In order to function as a scientist, you have to let go of the predictable. That has also given rise to more new developments. A science like Chaos Theory would have been unthinkable little more than a century ago.

Modern physics may be well above most of our heads but what it gives us ordinary mortals is freedom from the terrorism of number as we try to apply a scientific approach to understanding the soft sciences. For enterprise architecture it’s a godsend. It should be too for all the social sciences. It’s bizarre really that these sciences seem to anchor themselves in 19th century physics.

Instead of getting tied up in statistics and formulae, we can now concentrate on applying scientific method and also make use of the insights of the last 100 years. By scientific method I mean a process involving both the intuitive and the rational mind. I’ve quoted Einstein on this and since discovered other scientists with the same perspective. So measurement and understanding have a circular relationship with each other. The rational mind assembles all the information that is known about the system under consideration and about similar systems. The intuitive mind then takes over in a search for meaning in the information. When systems are complex and even more so when they’re chaotic, the intuitive mind plays a greater role. When systems are inherently amenable to simplification, the rational mind plays a greater role. In EA the latter comes into play when something (e.g. an IT system) is unnecessarily complex – more complex than the business model it’s supposed to implement. Let’s face it, there are enough examples of that, both in IT and in manual processes. But as I’ve said, most enterprises are of necessity complex and their architecture is therefore chaotic in the scientific sense that it’s highly sensitive to initial conditions. So a useful enterprise architecture needs to be able to respond to the unpredictable, which is of course what this whole series of blogs is about.

As Heisenberg pointed out, the equations for quantum mechanics also give correct results for (physical) systems at everyday scale. That’s what EA needs to achieve as well. We should not make things more complicated than they need to be but we shouldn’t make them more simple either than reality demands (and yes, I know there’s an Einstein quote for that one too).

Share

Tags: , , , , , , , ,

Comments (5)

Trackback URL | Comments RSS Feed

  1. Louis says:

    Great post Stuart! Just before I read it, I stopped for a short while thinking about the tweet of Jan Bommerez about systems: “As above, so below. As inner, so outer. It is all one system. The fragmentation is not in the system but in our minds.” This is a more or less spiritual notion of system, and for those open to it, probably very true.

    So when we start even talking or thinking of systems (or EA) in the “traditional” way, we are implicitly creating an illusionary fragmentation, which can (probably) never be truly or reliably measured, determined or scientifically proven looking only at it’s fragments (that we tend to call systems or structures or domains or architectures or …).

    So what is it that we humans so desperately want to control? Maybe it’s just an illusion…

  2. Peter Bakker says:

    Hi Stuart,

    Even after our tweet-exchange I’m still confused by your last paragraph.

    What is your call to action? Is it that EA should give the correct results by using the same scientific method on every scale? Or must EA use scientific methods to improve itself and the enterprise step by step (in the spirit of Vitruvius)? Or both, or something else?

    • Stuart Boardman says:

      Thanks Peter. I like this question, because it forces me to do what I always used to criticize one of my maths lecturers for not doing. I have to explain what I said in different words. So here goes.
      What I’m trying to say is definitely not your first option unless one were to interpret “the same scientific method” as “scientific method” in the generic sense, which is not what I think you meant. And since I have not yet read up on Vitruvius (I ought to remember what he said but I don’t), I’ll leave that one to my betters.
      In fact the last paragraph is just a short restatement of the previous paragraph. I believe that not every system can be deterministically described. The behavior of living systems and of some physical systems (e.g. weather, the North Atlantic ocean) is not deterministic. Any system involving living systems (e.g. an enterprise or a city or an ecosystem) is in general also not deterministic. The terms complex and/or chaotic are often applied to them. Neither a complex system nor a chaotic system (even a simple one) is deterministic. Simple and “complicated” (not my term) systems are in general deterministic. Some complicated systems are implemented in an unnecessarily complex way – not unusual in the world of IT. So “over-complex” (my term) systems should be amenable to simplification and may then be deterministic.
      Rationality and intuition are two parts of the scientific process. So a scientific method is one which applies an appropriate proportion of each of them to the solving of a problem. If we have a set of “equations” known to work for a complicated or simple problem domain, we can expect to be able to apply them repeatedly. For a complex or chaotic domain a set of “equations” known to work once cannot be expected to work a second time. They might but that itself is unpredictable. In this case and in particular for chaotic systems the degree of uniqueness (I’m not sure that’s good English but bear with me) is sufficient to reduce the chance of repeatability to negligible. This is where intuition plays a bigger role.
      In order to cope with this, we try to develop what we call “frameworks” to help us take all factors into account and to keep us honest – the right mix of intuition and rationalism. Sometimes frameworks are called methods or even methodologies. I don’t care. One has to understand them and experiment with them in order to know which will work best in any one situation (usually a combination). For more on this topic, you might want to follow the thread that begins with http://weblog.tetradian.com/2012/11/10/metaframeworks-in-practice-intro/
      Was that better?

      • Tom Graves says:

        I’d strongly agree with you here, Stuart: ‘scientific method’ as presented ‘for public consumption’ is very different (and, usually massively over-simplified) from ‘scientific method’ as actually practised. For the latter – which is really what we’re talking about here – see e.g. Beveridge’s ‘The Art of Scientific Investigation’: full-text available for free-download on Archive.org at http://archive.org/details/artofscientifici00beve

      • Peter Bakker says:

        It didn’t occur to me that the last paragraph was a restatement of the previous one because of the sentence “That’s what EA needs to achieve as well.”

        So I thought the last paragraph was some kind of addition…

Leave a Reply

Your email address will not be published. Required fields are marked *