Mix of technical, philosophical and pragmatic reasons:
1. Technical: My background is in control theory, the foundational field from which System Dynamics forked off, due to Forrester, in the 1960s. SD remains to this day “applied 1960s control theory.” Developments since then, in controls (40 years worth) make SD rather obsolete and flawed. That would be a longer, mathematical discussion that I don’t want to have on this blog’s comments sections :)
2. Philosophical: I think the SD folks’ basic philosophy of modeling is flawed, another discussion I don’t want to have here.
3. Pragmatic: Within the restricted domain of problems where SD is applicable and useful, four things need to be true for there to be value. i) the modeler asks the right question, ii) he/she has the aesthetics to build a model with the right level of coarseness/fineness depending on the question and the quality of data available (time constants etc.), iii) Has the plug-and-chug technical skills to code the model in a tool and run it without making conceptual mistakes, iv) Actually understands the plug-and-chug formulas properly to interpret the results right.
In my experience, people who have all 4 skills are generally smart enough to build a MUCH leaner model using a judiciously-selected mix of modeling tools that ends up being MORE expressive and answering the question better. So you get more for less. So the best modelers in my experience rarely pick SD as the right way to answer any interesting question.
On the flip side, if any of those conditions fail, you get dreck. The most common failure modes are a) asking unimportant questions that matter to nobody b) knowing only skill iii, plug-and-chug.
In the first case, you get good answers to questions nobody asked. In the second case, you get aesthetically ugly answers to the wrong questions, with the (correct) answers being interpreted incorrectly.
This is for classic system dynamics modelers who actually go to the trouble of even using the tools for questions answerable by simulation. There are 2 ways SD slides into even worse territory under the generic, meaningless label “Systems Thinking.” If I hear that phrase, I run a mile.
In the first way, some clever talker will use a single mind-candy example (like the bullwhip effect in the beer supply chain) to riff on huge, very different questions, wail about how people are idiots and end with some idiotic line like “to answer important questions like global warming, we really need to understand the real dynamics.” Such people absolutely don’t get how huge socio-cultural-political problems are ACTUALLY solved in the real world. The big variables are not the objective realities but interpersonal dynamics, personalities and other psycho-social factors playing out in the people involved in solving the problem.
The second way is even worse. Here, even the mind-candy examples (which at least reveal a single relevant insight) are dispensed with, and you get into purely philosophical territory about how the world ought to work. I couldn’t finish Peter Senge’s Fifth Discipline because I had objections on nearly every second page — it is a thoroughly shaky extrapolation from already shaky technical “systems thinking” foundations into organizational theory and self-improvement. But his book after that (forget the name) which is a series of conversations with a group of people who seem to want to change the world, was basically in New Age spirituality territory rather than science. That sealed the deal for me.
Hope that doesn’t offend anyone.
For the record, I do understand the technical end of SD, have played with the tools and have even had one paper in the IEEE Transactions on Systems, Cybernetics and Man :). . (It isn’t a core SD journal, cybernetics is the precursor created by Norbert Wiener, who was a contemporary of Vannevar Bush, Jay Forrester’s adviser). So this isn’t a completely unfounded opinion…
Still, after that long rant… yeah, basic stock-and-flow/iThink System Dynamics is a useful tool under some narrowly circumscribed conditions. It just isn’t the world-changing epistemological revolution its practitioners think it is.
I’d love you to go into this at some point, I’ve got very fond of the old-school cybernetics stuff, and one of my medium-term goals is to update it for the stuff that’s changed at the theoretical level in the meantime; chaos/topological stuff/computational mechanics etc, so that I can use it more effectively in a wider context.
Obviously that’s my project not yours, but I’d love to hear what parts of it you find useful and how that’s limited.
It’s a good idea for a project, but backward-looking IMO. You will unduly constrain yourself if you frame the new in terms of the old.
I think the most innovative thinking in this stuff is probably happening in 2 places: game design, and social technology design. I’d start there and theorize from first principles about basic assumptions, design laws etc. in those domains.