A characteristically smart and articulate post from George Siemens explaining why a view of the universe as nothing but networks all the way down – that he has supported in the past – is not sufficient to explain everything that matters. As George says, a systems view tends to be way more useful. It is important to observe that this is not in any way incommensurate with a network-oriented view because systems are entirely about networks, network theories play a very important role in modelling and understanding systems and, in fact, network theories are just a subset of systems theories anyway so, as George points out in this essay, he was not actually wrong in the past. It’s just that (perhaps – I present a counter view at the end) he could have been more right.
Not just one theory but many
There’s a great deal of diversity in systems theories, crossing many disciplinary areas, with different standards for rigour and explanatory power, and that’s part of their strength. They offer ways of talking about systems that are appropriate to their context. What is common to all systems theories is that they are anti-reductive, focused on relationships and interactions between things over time more than their constitutive elements, but there’s a host of different ways that broad approach can be applied. Personally, I am particularly drawn to the field of self-organizing systems, which means an interest in the general areas of cybernetics, complex adaptive systems, autopoietic systems, signal/boundary systems, evolution, stigmergy, swarm intelligence, networks, etc, but there’s a lot of other helpful kinds of system theory. I have found Michael Moore’s much higher-level systems view of education, for example, to be really useful in my research and teaching, and approaches like systems dynamics can be very helpful to understand why systems that surround us constantly fail. Systems models can, for instance, help to explain why incentive systems reduce motivation, or how -more generally – systems (once created) develop their own goals independently of and often in direct opposition to the people within them or their creators. Systems views are not always presented as such. One of the most life-changing books I have ever read, for instance, is Jane Jacobs’s The Death and Life of Great American Cities, which presents a rich and poetic systems view of what makes a city area thrive or fail, and has been hugely influential in driving the development of many cities around the world, though there’s far more to it than that. There’s barely a network to be found within it, and it doesn’t draw on any formal systems theories, but it certainly contains one. Though others have developed more network-oriented systems theories out of it (notably Christopher Alexander and Bill Hillier) the power of Jacobs’s systems theory is far more to do with the richness of her storytelling and her complex, multi-layered, deeply humane analysis of human systems. The level of detailed observation and depth of insight is similar in many ways to that of Charles Darwin, another preeminent and seminal systems thinker who did not label himself as such. Both Darwin and Jacobs do not simply show that – they show why and how, in wondrous and complex detail, everything affects everything else. Some systems can be useful but rather boring, especially when they are closed. Computers, for instance, are systems of interoperating parts and layers. They are complicated, for sure, but not (in themselves) complex. This makes them, as systems in themselves, a bit dull. Sitting by itself, notwithstanding interesting ways it can be programmed to adapt, a computer is essentially a closed system that behaves in predictable ways. However, the field of information systems is much more about human systems than computers, the field of computing as a whole is rich in invention, and the field of software development using computers is fully open and truly complex, full of unexpected and emergent behaviours, combining ideas, fields, groups, individuals, and models from all over the place. Connected together, they can do very interesting and sometimes unexpected things. Computers are (mostly) boring systems, but they are part of, and are used to enact or contain many much richer systems. Similar things can be said of legal systems, accounting systems, most machines, many organizations, and so on. It’s true of many systems in nature, too, such as metabolic pathways or neural connections. In themselves, they are (I simplify a little) systems of interacting mechanical processes following a set of simple rules. Things only get really interesting when you look at them as subsystems of other systems, interacting with other subsystems, whether creating something planned or emergent. Of course, it’s not just about things with lots of parts. Even simple, uncomplicated systems can be complex: the classic three body problem is a good illustration of this. It’s about how those parts are configured, and their openness to energy or information from the environment.
More is different
Systems theories that go beyond mere networks are necessary because more is different, as P.W Anderson famously demonstrated way back in 1972, and new laws, principles, patterns, and concerns occur at many different scales. Such laws and regularities are inherently unpredictable from the behaviour of their parts (see Kauffman’s Reinventing the Sacred or Humanity in a Creative Universe or even his older Investigations for a solid theoretical explanation of why this must be – it’s all about adjacent possibles) so, even if you can posit a theory that consists of networks from bottom to top, there’s limited value to be gained from doing so. It’s like string theory – if true, it probably does explain nearly everything in the whole universe but it’s not a lot of help with your shopping or filing your tax returns. Network theory strips a lot of what is meaningful from the system it models. There is a great deal that can be learned about learning from an understanding of the dynamics of networks, but they are of limited value in helping you to, say, construct a learning plan for yourself or others, or figure out why you are procrastinating about your homework right now.
Signals and boundaries
I tweeted a rather opaque response to George’s announcement of his article, in which I mentioned signals and boundaries. That’s worth unpicking a little. The central concept comes from John Holland’s brilliant eponymous (and, sadly, last) book, Signals and Boundaries. For any system that we choose to look at, we must choose which are the boundaries that matter to us, examine the signals that pass between what is at either side of those boundaries, and consider what tranformations occur within the boundaries (not necessarily how they occur), in order to understand it at an appropriate level. Though there are some consistent patterns at every scale (that Holland brilliantly reveals) we come to very different understandings depending on the boundaries we choose: the rules, the signals, the behaviour of the systems, etc are, qualitatively, profoundly different. For instance, consider the difference between anatomy and metabolic pathways in cells. You can’t have the former without the latter, but there is no conceivable way you could deduce the function or form of the heart by looking at enzymes in cells (of course, you could learn useful things about how the heart works by looking at metabolic pathways because they are subsystems or, a little more accurately, sub-sub-sub-subsystems of the heart). Choosing boundaries is a process of black-boxing wherein, once a significant boundary is chosen, we treat the internal part as a kind of ‘program’ that processes the signals it receives and evokes responses. This is what I think George is getting at when he suggests that what makes systems different is that they embody rules. This is smarter than a simpler network view in a variety of ways. It makes it easier to focus on levels that matter, using context-appropriate vocabularies and meanings, in whatever combinations are significant; it allows us to more easily combine different scales/granularities of boundaried entity; it allows us to think more deeply about qualitative as well as quantitative differences in the signals; it allows us to think about not just networks but sets, or organizational structures, or whatever is appropriate; and (arguably most usefully) it makes it far simpler to think about processes (the ‘programs’) that drive it, and how they affect one another. It does all this without losing any of the value of looking at it as a network.
Connectivism as a systems theory
In fact, though George is a little dismissive of his most famous and widely cited article on the subject, a lot of this kind of systems perspective appears within it. He talks of ecologies (archetypal systems) quite a bit, explicitly mentions systems theories as playing a foundational role in setting the agenda for the theory he expounds, spends a fair bit of time on chaos theory and self-organization (both explicitly systems fields involving systems theories), and even, as he discusses the implications towards the end, explicitly refers to connectivism as “a systems view of learning”. Though not explicitly mentioned, the theory also draws quite a bit on the field of socially distributed cognition, which is essentially a systems view of knowledge. So, though George may have meandered off the path a bit along the way and got caught up in trying to make everything look like a network from time to time, the version of Connectivism that most people adopt is based on this paper, which is and has always been about a systems theory, rather than a network theory. Even its central message supports this view. Of the eight most oft-quoted principles at the centre of the essay, only three are explicitly about connections. The rest are concerned with processes, axioms, and attitudes that relate to what’s inside the black boxes (the network nodes). These might, charitably, be seen as supportive of networks, but are far more to do with how to learn in and as part of a self-organizing complex adaptive system rather than how the network itself embodies learning. That’s a big part of what makes it useful: we need such theories to make sense of the changing context in which we find ourselves, in which older theories (especially those embedded in a view of education as a formal process involving a teacher) seem inadequate. It also prevents it from being a complete theory of learning – there are other theories and models that take a different systems view (or even, perhaps, a non-systems view) that may be more appropriate, at least in combination with it, to some circumstances – but that’s no bad thing. In fact, it is kind of implied in one of the central axioms of the theory itself: “Learning and knowledge rests in diversity of opinions“. This has been one of mine.
Originally posted at: https://landing.athabascau.ca/bookmarks/view/4119883/george-siemens-says-he-was-wrong-about-networks-well-not-exactly-wrong