I came across stocks and flows a couple of years back and it certainly changed my thinking. Interestingly this is something we sorta commonly use in backend dashboards to understand the behavior of the system, especially with regards to data. I always thought there ought to be some similar ways to envision economy with this and I find infographics trying to bridge the gap albeit still static though.
You may be interested in reading more about system dynamics. There are a number of good books out there:
- Thinking in Systems, also by Meadows
- Business Dynamics, by Sterman (slowly working through this)
- Strategic Modelling and Business Dynamics, by Morecroft
- General Systems Thinking, by Weinberg (on Leanpub you can get the PDFs, it's 4 books there, as I recall the first two were published as one book when it was published as a paper book)
I've found Business Dynamics by Sterman to be the best all-round System Dynamics book in my collection (I have three of the above, excluding Weinberg's).
Meadows is interesting and a quick read, but shallow. Morecroft is a thorough read and a decent alternative to Sterman's if you can't get it. But Sterman's really is the best textbook.
I got my copy of Sterman from my sister (who had studied systems engineering at MIT and so had a copy from her coursework there), but at your recommendation from a few years ago.
I worked through (most of) Morecroft because it was available through O'Reilly's online library and I had a subscription through work. I read Meadows specifically because of the leverage points essay, and wanted to see what else she had. It's definitely shallower than the others, but I think a good introduction because of its relative brevity.
Ashby's book is pretty decent, considering that Cybernetics as a field can feel unapproachable to lay folk. In terms of DNA, cybernetics and systems dynamics both draw deeply from the well of early control 20th-century theory and so there are some conceptual overlaps. In relative terms I think cybernetics does a better job of surfacing the contributions of information theory and system dynamics does a better job of surfacing the impact of stocks and flows on overall dynamics (as its name would suggest).
Oh drat, they used to have free examples you could play with. (Edit: they still have a free plan, after signup, so I still intend to try it)
I've seen Loopy before, but Sheetless looked more like existing tools with storied histories (Stella etc).
Like any forehead-smacking moron who can code I have sometimes considered writing my own because the thing I miss most of all in systems dynamics tools is being able to use git and write tests. But I know now that writing real software is really hard.
This post by Will Larson (author of the engineering management book, An Elegant Puzzle, which was published by Stripe Press) mentions some tools for system modeling with feedback loops: https://lethain.com/systems-thinking/.
Good article (or so I've heard, Leverage Points is still on my reading list) but worth noting that Robin Sloan is talking about a different sort of stock and flow. His usage is more akin to long-form evergreen media versus in-the-moment short-form content (e.g., tweets) and striking a balance between the two.
I've read a lot of "system analysis" / "system dynamics" literature in grad school, and did workshops with practitioners, and my impression is that they suffer a lot from "if you have a hammer".
Business people and ecologists discovered modeling with ordinary differential equations, created their own jargon and tools around it, and decided it is the key to understand everything.
ODEs are a great tool to have in your modelling toolbox, but they're not, like, revolutionary.
To give credit where it's due, they manage to explain simple differential equation systems in accessible terms, and they manage to get people who would otherwise not build, test and think about explicit models at all, to do it to some degree.
But if you're already someone used to thinking with models, this is 101 stuff. My gripe with the discipline is that they barely go beyond the 101 stuff and hype their methodology to high heaven. The field is also permeated by relativist philosophy, or at least, it was 5 years ago when I was keeping up with the System Dynamics Society.
It leaves you in an awkward spot as a discipline if your bread and butter is building models, but you don't think there's an objective truth to test your models against, since truth is just group consensus.
I think it's fair to say that System Dynamics has two major strands of thinking. One grows out of the original Forrester style -- managerialist, looking for hard numbers, wanting to solve hard problems. I count Sterman in this strand; he has himself lamented that SD as a field has not kept up with developments in other fields it could adopt[1].
The second strand is more woo-ey. Meadows falls into that category. It's fun, but I don't use SD as a thinking harness because I want to be one with anything or achieve Satori. I use it because I am too stupid to solve systems of differential and integral equations merely by squinting at things.
I quite like DES as a style too. It's tremendously more efficient and easier to debug. I sorta-kinda smooshed them together in a simulator I built[0].
25 comments
[ 2.9 ms ] story [ 54.0 ms ] threadhttps://en.wikipedia.org/wiki/DYNAMO_(programming_language)
I came across stocks and flows a couple of years back and it certainly changed my thinking. Interestingly this is something we sorta commonly use in backend dashboards to understand the behavior of the system, especially with regards to data. I always thought there ought to be some similar ways to envision economy with this and I find infographics trying to bridge the gap albeit still static though.
- Thinking in Systems, also by Meadows
- Business Dynamics, by Sterman (slowly working through this)
- Strategic Modelling and Business Dynamics, by Morecroft
- General Systems Thinking, by Weinberg (on Leanpub you can get the PDFs, it's 4 books there, as I recall the first two were published as one book when it was published as a paper book)
Meadows is interesting and a quick read, but shallow. Morecroft is a thorough read and a decent alternative to Sterman's if you can't get it. But Sterman's really is the best textbook.
I worked through (most of) Morecroft because it was available through O'Reilly's online library and I had a subscription through work. I read Meadows specifically because of the leverage points essay, and wanted to see what else she had. It's definitely shallower than the others, but I think a good introduction because of its relative brevity.
http://pespmc1.vub.ac.be/books/IntroCyb.pdf
I've seen Loopy before, but Sheetless looked more like existing tools with storied histories (Stella etc).
Like any forehead-smacking moron who can code I have sometimes considered writing my own because the thing I miss most of all in systems dynamics tools is being able to use git and write tests. But I know now that writing real software is really hard.
Might be interesting to you based on the above.
Business people and ecologists discovered modeling with ordinary differential equations, created their own jargon and tools around it, and decided it is the key to understand everything.
ODEs are a great tool to have in your modelling toolbox, but they're not, like, revolutionary.
To give credit where it's due, they manage to explain simple differential equation systems in accessible terms, and they manage to get people who would otherwise not build, test and think about explicit models at all, to do it to some degree.
But if you're already someone used to thinking with models, this is 101 stuff. My gripe with the discipline is that they barely go beyond the 101 stuff and hype their methodology to high heaven. The field is also permeated by relativist philosophy, or at least, it was 5 years ago when I was keeping up with the System Dynamics Society.
It leaves you in an awkward spot as a discipline if your bread and butter is building models, but you don't think there's an objective truth to test your models against, since truth is just group consensus.
The second strand is more woo-ey. Meadows falls into that category. It's fun, but I don't use SD as a thinking harness because I want to be one with anything or achieve Satori. I use it because I am too stupid to solve systems of differential and integral equations merely by squinting at things.
I quite like DES as a style too. It's tremendously more efficient and easier to debug. I sorta-kinda smooshed them together in a simulator I built[0].
[0] https://github.com/pivotal/skenario
[1] "System Dynamics at Sixty" in System Dynamics Review Jan-June 2018. There are some PDFs around: https://scholar.google.com/scholar?cluster=15699150741004493...
There are definitely useful tools there, I think it's great to make use of this stuff, as long as you don't drink the kool-aid.
In this piece, he references the Garden and Stream concepts, while also adding in his own, Campfires. https://tomcritchlow.com/2018/10/10/of-gardens-and-wikis/
Then in this follow on piece, he touches on Stock and Flow referencing the original article. https://tomcritchlow.com/2019/02/17/building-digital-garden/