antlerboy - Benjamin P Taylor 7:44 am on May 15, 2021

An introduction to complex system science – Roli (2015)

An introduction to complex system science ∗
[PDF] An introduction to complex system science ∗ | Semantic Scholar

An introduction to complex system science

A. Roli
Published 2015

∗I am grateful to Prof. Silvana Bettelli Biolchini, who disclosed to me the beauty of science and introduced me to fractals, chaos and dynamical systems when I was a student at the high school. I would like to dedicate her these notes, hoping that she will enjoy reading them, remembering those pioneer times in which we had to wait a whole day to get the first fractal image on a computer monitor. campus.unibo.itSave to LibraryCreate AlertCiteLaunch Research FeedShare This Paper

antlerboy - Benjamin P Taylor 5:36 pm on July 4, 2021
Good read. Thank you for sharing.
I thought you don't believe in the computability of the world😁

— The Kihbernetics Institute (@Kihbernetics) July 4, 2021

I accidentally replied to a tweet about this with “the opening is tragically confused, distinguishing ‘the system’ from ‘the model of the system’ – the reductionism at the heart of so much SFI-flavoured ‘complexity’ is fascinating.”

I didn’t mean to be controversial but it was a public comment and Michael Garfield politely asked about it so I feel duty-bound to reply. I should specify that I will respond about the piece. Including the Santa Fe Institute was probably a mistake – but there is certainly a scientific/mathematical approach to complex, including much agent-based modelling which fits with the pattern I identified in the paper (I’ll return to this briefly at the end).

To give the lede up front:
- I believe that scientific study ‘of systems’ is valuable, that looking for mathematical patterns can be very valuable, that patternicity is interesting (I think that underlying laws – slightly different, and I don’t actually see a focus on them always emerging from looking for patterns)
- this tends to lead to a kind of naive realism which sits uncomfortably with the constructivism at the heart of the discipline; it’s really the failure to robustly acknowledge this tension which I think is the gap
- this failure is at the heart of the tendency to talk about ‘complex systems’ in the world (rather than ways of seeing),
Somewhat ironically, I think that it would, on the whole, be better to have clearer definition and separation between study, theory, philosophy, and practice in this field, and between physical, ecological, and social systems (for example). It’s (obviously) not that I think that these are completely distinct spaces; it’s more that it would reduce the tendency for a failure to address these issues and distinctions leads to an unconscious elision of categories which leads to, for example, valid work in the analysis of physical or ecological systems being taken as valid theory for social.

First of all:
‘Complex system science (CSS) is a fairly recent field of science’
Well, my opinion is different to this, so of course I’m off to a difficult start (https://stream.syscoi.com/2020/04/21/bringing-together-some-reason-and-old-threads-on-systemsthinking-is-complexity-is-cybernetics/)

So from a ‘this is a science perspective’, we start with ‘There is no formal definition of a complex system— and it is unlikely that a formal definition indeed exists’.
This might just be slightly loose drafting (or translation – there are one or two lines which suggest this – but I wonder if this is doesn’t hint at the locus being placed in the world not the ‘corpus of methods’. Of course, it’s not surprising that in an introductory teaching piece there’s a search for definitions, but it’s notable that it balances the lack of definition with a search for it in the world.

(And the list of properties raises some interesting questions for me – I recognise that this is just a suggestive sketch of features, but there’s a mix of basic descriptive elements, analytic categories or approaches, and emergent properties which I find intriguing! And ‘robust’ seems to be… debatable… at least?)

“CSS has the objective of studying and modelling, and controlling CSs. It is interdisciplinary and it involves many disciplines, such as mathematics, physics, computer science, biology, economy, philosophy, neurology and more.”

Fascinating to me that this is precisely the ‘damning charge’ levelled at systems thinking and cybernetics, to which ‘complexity theory’ is often contrasted! Hey ho. But I take this to be an honest and accurate definition of this particular space. The word which is evident by its absence is ‘understanding’. This is first-order black box thinking at its’ finest. (And, of course, potentially hugely useful – except that you must be clear that the framing of the black box determines your understanding of it and the implications thereof).

“A peculiarity of complex systems is that local rules often propagate information in such a way that the whole system is subject to a dynamics that is not possible to understand solely on the basis of the system’s constituents. Indeed, it is often said that some global behaviour emerge in a complex system from the interaction of its parts, governed by local rules. Despite its relevance for CSs, the concept of emergence remains somehow elusive. I quote below three definitions that provide a sufficiently clear framework for this phenomenon.”

I just want to note that I appreciate the approach to definition again here – examples in practice and use, rather than one ‘formal’ static definition.
And also that ‘local rules often propagate information’ is an interesting turn of phrase – it suggests to me an assumption of how the world works, or the application of a way of doing science onto the world itself (which come to the same thing); this is, for me, the wrong way around.
To be specific, a rules-based modelling approach is very useful. But to assume that because it is useful, this is an appropriate way to think of how the world works is (a) to misunderstand the scientific method (missing the second-order and critical perspectives, epistemological humility, and methodological scepticism) and (b) to go looking for trouble when it turns out this isn’t an appropriate way to think of how the world works.

By the way, the Yanneer Bar-Yam definition of ’emergence’ is particularly provocative and interesting:
“Emergence refers to all the properties that we assign to a system that are really properties of the relationship between a system and its environment.” (from Y. Bar-Yam, Concepts in Complex Systems, 2000. Available on the New England Complex Systems Institute website, http://necsi.edu/guide/concepts/emergence.html)
This URL now resolves to https://necsi.edu/emergence where I can’t find that precise quote, but I can find:
“One of the problems in thinking about complex systems is that we often assign properties to a system that are actually properties of a relationship between the system and its environment. We do this for simplicity, because when the environment does not change, we need only describe the system, and not the environment, in order to describe the relationship. The relationship is often implicit in how we describe the system.”
I don’t have time to do this justice now, but might come back to it later; this seems very insightful but ‘assigning qualities to the system’ is not something we should properly be doing anyway – we are always talking about behaviour and its’ meaning in a context, and that is the mistake that I’m trying to winkle out here. It strikes me as half-way to the real insight.
Also, the use of ‘qualities’ and ‘properties’ rather than ‘descriptions’ again points to a kind of reification.

Next quote: “This is the case of the word system, which is usually referred to objects, things or entities (both concrete and abstract) for which it is possible to identify some kind of boundaries so that the system can be distinguished from its outside. (Of course, this is a generic definition) Therefore, once we have identified a system, it is possible to say what belongs to the system and what does not, i.e., what is outside of the system’s boundaries (its environment). The identification of a system depends upon the specific level of abstraction chosen and the viewpoint of the observer. Usually, we can also say that a system is composed of other entities, whose relationships are more strong than those with the outside of the system.”
As with the Bar-Yam piece, the iteration between scales is really valuable and insightful. And an understanding that the observer and observation are inherently part of the description of the system is advanced.
But ‘possible to identify some kind of boundaries so that the system can be distinguished from its outside’ is the reification of the concept that there are ‘systems’ and ‘boundaries’ in the world. We don’t ‘identify’ a system, we ‘define’ a system.

To be precise: to do science, we must give ourselves the right to say ‘this is the thing’; we only *earn* that right, however, to the extent that we are clear that this is a *decision* and to the extent that we *own* that decision (more or less arbitrary and always, ultimately, non-rationally founded, but necessary nonetheless) and its implications.
This is my challenge in communicating with (systems) engineers and (complex) systems scientists, and it’s a small but important thing. It’s the people who are embedded in paradigms or assumptions versus the people who understand, at the very least, that they are not inherent or eternal – nor are they meaningless – but they are what create meaning).

There’s something of this in there, to be sure, with “The identification of a system depends upon the specific level of abstraction chosen and the viewpoint of the observer.” – but passed over so quickly that it’s not even a kissing cousin to the constructivist reality. This is reductivist because it implies there’s a ‘right answer’ for systems boundaries – one which is dependent on level of abstraction and viewpoint, but dependent in the way that those trick figures which can cast shadows that are circular, square, triangular etc are dependent – it is assuming that the ‘systemness’ is out there, to be grasped.

It might seem, so far, as if I am really complaining about realism versus constructivism, and perhaps that is the case. Look at:
“The identification of systems is the first step into the study of the world around us; indeed, ancient philosophies and biology are full of discussions on systems and their classification and taxonomies.”
This seems to be a continuation, not a critique, of classification and taxonomy, which I believe is a trap into which we can easily fall (at this point I’m abandoning ‘complexity science can fall’ because there are really no distinct propositions properly boundaried by that phrase; complexity people and science and work can fill all the spaces on my map, as can that branded ‘systems’ or ‘cybernetics’ or ‘practice’ rather than science etc; which is why those tags are not helpful).
But the reductionism, for me, comes from the modelling relation which then comes from an assumption that there are ‘systems in the world’ and our task is merely to get their classification, naming, and rules *right*’.

“However, another concept is needed to systems’ science: the model of a system. A model is an abstract and schematic representation of a system. It is often a formal representation of the system. It has to be emphasised that a model represents a portion of the system, as it captures only some of its features. In addition, it requires an abstraction process, which involves simplification, aggregation and omission of details. In general, we make a model of a system to (i) understand it and investigate some properties, (ii) control it, and (iii) make predictions on its future. Important classes of models are differential equations and difference equations. However, in general a model can be any system we decide to use to represent another system. For example, we may want to model a biological system (e.g., a cell) by means of a computer simulation or of a wet lab experiment.”
It seems clear to me here. We have ‘representation of a system’ and ‘captures only some of its features’ and ‘abstraction process, which involves simplification, aggregation and omission of details’. Yes, yes, and yes – all that’s missing is ‘it requires a formalism which creates an interpretation of our understanding of the system from a particular perspective, for a particular purpose, in a particular tradition and framing – and remember that we boundaried the system for a purpose, in a particular tradition and framing, in the first place’.

Professor Mike Jackson put it well in his recent discussion on complexity (I’m paraphrasing broadly) – ‘yes, it’s not surprising that when you abstract from a human system all those aspects of humanness, it looks very comparable to a non-human system’. Science is shadow-boxing; interrogating our understanding of reality from within our understanding of reality. This makes its insights doubly miraculous. I worry that a lot of model-based complexity science is accidentally tautological; seeing its own framing and therefore celebrating its victories.

I am not advocating, here, for solipsism or relativism. That there is a world is the only practical way to proceed, and that we make genuine discoveries through science is undeniable and exciting. If ‘the map is not the territory’ is valuable, then ‘the only thing you can compare the map to is another map’ is equally or even more valuable. Testing propositions in the real world allows us to reframe our perceptions and improve the meaning-making and practical application of our work for a purpose, within a frame. It just does not allow us to say that our new map is closer to the territory in any sense that means more than that.

In this context, the definition of ‘types of systems’ (while still potentially useful) is an exercise in classification and taxonomy which says as much about the classifiers as it does about the world (as always). I was tickled by ‘dynamical systems – those that evolve over time’, which suggests that other types of systems, perhaps, do not.

So, for me, the reductionism is in ‘systems in the world have actual boundaries’ which leads to believing that we are comparing our modelling to the world when we are comparing our modelling to our other modelling, and (as shown by reference to ant colonies and the like) in taking things out of their meaning-giving frame then claiming that we know them, because we recognise what is left.

In my view the real insights are:
- to take a critical perspective to the frames we must necessarily place on things (a move to a second-order perspective)
- to therefore think about potential reframings rather than simple application of framings that worked well elsewhere
- to get inside the meaning of things at the level at which the meaning exists, rather than above or below, with insights from the levels above and below
We started off with the claim that ‘complex systems science’ is new. I also want to say that the discussion about ‘science’ is always a tricky one for me, I respect science and scientists and the scientific method; in the systems/complexity/cybernetics field, I would say that there’s real value in the rigour and progressiveness of science and real risks in a reversion to ‘disciplinarity’ in an inherently transdisciplinary field.

So all the learning outlined in the piece I’m critiquing is potentially hugely valuable, and even if mistakenly reductionist, it can still produce practical results and affordances to insight. The risk of bringing in the constructivist perspective is that it leads to nihilism, solipsism, or relativism. The risk of not doing so is short-sightedness and reductionism.

(PS It may be worth, if you are interested, comparing to my little piece ‘complexity is not a paradigm shift’ – which, ironically, was aimed at a *different* ‘complexity’, an approach to complexity in the social sciences which paints systems thinking as reducationist and control-focused: https://chosen-path.org/2021/05/20/complexity-is-not-a-paradigm-shift/ and which advocates the true insights as:
1- seeing the organisational potential within the system
(from whatever source or level – systems/complexity/cybernetics can equally identify/support/set up structured, command-and-control hierarchies (say, Jacques’ Stratified Systems Theory), and the conditions for ‘pure self-organisation’)
2- seeing the role of the observer, a second order perspective
that the observer creates the ‘system’ (or the complexity or chaos); that talking about these terms – or pattern and nebulosity in general – outside of a loop between the observer and what is being observed – is meaningless
3- making critical boundary judgements, i.e. understanding that the framing we bring to our observations has implications (this follows from (2)) )
PY 8:39 pm on July 4, 2021

Thanks for the clarification. Just a couple quick points:
1. I agree with you that science was dealing with complexity long before the “invention” of CSS. In fact I will go further and claim that all of science is an attempt to reduce the perceived complexity of the world. Yes, perceived, because I strongly believe “complexity” is in the “eyes of the beholder” (observer). Everything seems complex when you don’t know a thing about it.
2. That being said, CSS is a “new scientific discipline” because it emerged only recently from a perception (again) that traditional science disciplines cannot deal with complexity in a satisfactory manner. What I see, however, and the content of the “teaching material” at the centre of this discussion clearly shows, CSS brings nothing new to the table.
3. I think models are distinct from and a necessary element in the understanding of both, the system and the “machine” (phenomena) they represent. In fact, I think your “attitude” towards modelling may be the source of your perceived “discomfort” with systems engineering. All we do in engineering is building models (products) to share with anyone who needs them. And they work (well, most of the time). We are all aware that ‘the map is not the territory’ but, nevertheless, we use it regularly to navigate (compare with) the territory. To ‘compare the map to another map’ is just to find out which one is a “better” (more useful) representation of the territory.
antlerboy - Benjamin P Taylor 7:20 pm on July 8, 2021

No problem with any of that, and no problem with modelling at all as long as the partial perspective (etc) of the model is kept relatively visible. My frustration with engineers tends to be having to explain to them that no map is the territory…
- PY 8:55 am on July 9, 2021
  
  Yes, we are all human, and sometimes get so “enamoured” with what we do that we think that’s nothing else. I too had to have quite a few discussions with my SW engineering friends, because they thought they are building “the system” by developing software. I might have convinced few of them that they are not, that the “system” is already out there and people are already using it. All we do is building some new products (and changing associated processes) that are hopefully better and will replace the old ones. One important thing though is that we have to spend some time in showing (convincing) people that the new way of doing things will benefit them.

Systems Community of Inquiry

An open discussion on systems

An introduction to complex system science – Roli (2015)

An introduction to complex system science