Loopy by Nicky Case


New from Nicky Case is Loopy, “a tool for thinking in systems.” From the page:

In a world filled with ever-more-complex technological, sociological, ecological, political & economic systems… a tool to make interactive simulations may not be that much help. But it can certainly try.

Essentially it’s a way to build systems diagrams, playtest them out, and share them with others. Additionally, you can remix the work of others, too. It’s a collaborative system editing system? It’s a lot of fun to play around with.

See also Nicky’s interactive post mortem / making of on Loopy.


This is really neat! I’ve seen little GIFs bounce around on Twitter, but hadn’t realized that this thing had shipped :slight_smile:

When I was in school, I TA’d a course in Discrete Simulation that used a piece of software called SIGMA - the UI of Loopy reminds me of that software (though: SIGMA was written a bajillion years ago, and Loopy is much more playful / accessible!)

We used SIGMA to model complex systems (for example, a variant of it is used by Bio-G to model biopharmaceutical operations and evaluate alternate supply-chain variations) but the models in-class were modeling simple carwashes, restaurants, movie theaters, etc - you’d set up an arrival schedule, events would occur on the nodes, which would under-certain-logical-cases trigger other events to happen in the future.

I love how simple Loopy is - though I could also see unlocking logic-statements (“this relationship only applies if the value in the node is > 2”, or “this relationship only applies if some other value is zero” etc) to model state-changes.

(That said, it’s also wonderful as-is!)


i was pretty intrigued by loopy and played with it a few times but i think the part where i was less sure about it was that it was too easy to game a feedback loop to exhibit the behavior i wanted. I kept thinking about part 2 of the all watched over by machines of loving grace and the sort of tragic fallacy of using systems thinking to formalize what end up being complex and overtly human interactions that don’t behave in easy-to-model ways.

namely, i mean, check this part of the wikipedia synopsis (linked above)

At the time, there was a general belief in the stability of natural systems. However, cracks started to appear when a study was made of the predator-prey relationship of wolf and elks. It was found that wild population swings had occurred over centuries. Other studies then found huge variations, and a significant lack of homeostasis in natural systems. George Van Dyne then tried to build a computer model to try to simulate a complete ecosystem based on extensive real-world data, to show how the stability of natural systems actually worked. To his surprise, the computer model did not stabilize like the Odums’ electrical model had. The reason for this lack of stabilization was that he had used extensive data which more accurately reflected reality, whereas the Odums and other ecologists had “ruthlessly simplified nature.” The scientific idea had thus been shown to fail, but the popular idea remained in currency, and even grew as it apparently offered the possibility of a new egalitarian world order.

qv, an image from the postmortem:

i think in part what felt unsatisfying is that some of these feedback loops don’t actually respond linearly. like, i think about how, in practice i react to my phone showing me an arbitrary “sign in to icloud” prompt. the first time i see it and i’m like wtf is this nonsense and subsequent times i’m still horrified but just a little less. and maybe my fall-off is linear or maybe it’s logarithmic or maybe one day i just do it to show my kid a video we paid for and it’s like a unit delta for me not caring anymore. how do you model that? do you? and if so, what’s the point other than the more trivial “people hate being pesterd until you eventually break their will.” do you need a fancy diagram for that? and i mean, is my experience really relevant or indicative of the larger water torture of being forced to constantly reauthorize my device?

that said, the tool itself is quite nice (even if the ui is a bit opaque at times) but i sort of worry that the takeaway from people misrepresents the nuance, complexity and overall limitations of systems in general. like, how do you even calibrate a loopy diagram? how do you know that you’ve set the outputs from your root nodes correctly? and if so, why is it that the world often behaves counter to what you observe?

i think that on the whole, it’s a solid entry, but, like an episode of serial, “raises more questions than it answers”


Sigma looks so cool (and thorough)!! Did you find it was a good learning experience? I’ve never heard of a simulation class but I’m glad it exists!


I think it’s a good start, but you bring up a lot of great questions. One of the chapters in the thesis I recently read dealt with systems simulation, in particular cities (and their social systems). These were the systems which eventually inspired Sim City, it turns out.

What’s interesting about those early simulations is that while they’re much more detailed than a lot of the run-of-the-mill systems examples we see (like the simplified examples given by Loopy), they are nonetheless still models. As with all models, it’s not possible to simulate everything, only a subset / abstraction of “reality.” That simulation may not be 100% faithful, but it often can at least be useful, for the specific thing you’re attempting to illustrate or understand. The problem arises when the simulation stands in for the real thing in people’s minds (the map is not the territory?).


I really enjoyed learning SIGMA, yeah! I had a walk with my prof at the end of school, where I asked “ok, so… now we’ve spent the last couple of years learning how to make these simulations, what do I do with this in work?” The answer was something along the lines of “Well, simulations are used to make Big Important Decisions, and while you won’t be the person in charge of making those decisions today, you might be in that seat in 5-10 years - and if/when you are, you’ll know when to be skeptical of a modeling decision, and you’ll know what questions to ask of the person who made the simulation.”

A frequent theme throughout the classes was that a simulation is only as good as its assumptions. So it’s neat if you have a simulation that looks cool (many tools have extensive 3D visualization bits to 'em, like making a factory floor look real) but if those tools have crummy assumptions underneath them (for example, exponential interarrival times*) then the results aren’t very useful for the purpose of making a decision.

It’s a tough balance - if the tool looks flashy, people want to explore it, or the modeler might have an easier time pitching non-technical peeps on their consulting services. However, the results that come out of flashy/approachable tools may not have accurate assumptions, so they’re not as useful for making critical decisions.

(Of course, Loopy isn’t about making critical decisions, it’s about playing with feedback systems, so it’s a great fit!)

*exponential interarrival times are usually a sign that a simulation has made poor assumptions. They have some weird properties, like a nonzero chance that no customer ever visits your store again, and they have a memoryless property that works sometimes but not always. Modelers often use it because it has one variable, so it plays nicely in spreadsheets and feels approachable.


This is so important! And I’ve been feeling this a lot lately. There’s kinda two things being smushed together here:

  1. It’s a model and thus like you say, there are assumptions baked in. So it’s good to be aware of these assumptions. This is making me wonder, is there a way to somehow encode + show your assumptions in a model, such that they’re transparent and that you can do things with them? (I don’t know exactly what you could do with them beyond being aware of them, but an assumption-as-model-itself seems like a good jumping off place for exploring new models in the simulation medium).

  2. The other thing is kind of like media literacy (in a broad sense of “media” but also in the high school Media Studies-style school class). Media literacy as in being able to ask and answer questions like what are the assumptions here? and who is producing this? why are they making it? what are they trying to convince you of? and what’s in it for them to do so? That sort of thing. And again, the more transparent our media can be, the better equipped we are to ask and answer these sorts of questions, but I don’t know of many media which make these sorts of questions an important part of the discourse.