Loopy: a tool for thinking in systems | Hacker News Reader

Loopy: a tool for thinking in systems

ncase.me

508 pointsbpierre84 comments

I found a solution to the 3rd example problem on Automation and Unemployment : http://bit.ly/2nbHqX3

Much easier to manage than UBI. I see a bright future ahead.

ED-209 wants to have a word with Robocop + I decreased frustration and societal unrest by massively funding Entertainment:

http://bit.ly/2mwwO7T

Basically, "Running Man" was entertainment / fiction when I read it as a kid, and it will be a documentary by the time I'm dead.

I love it! Now if we can put everyone to work as extras in all that entertainment... Clearly we need to encourage automation in the real economy and ban it in entertainment.

This tool does what it says: the conversation goes up a notch because the feedback effects are part of the shared context so you don't get distracted talking about them one-by-one to the exclusion of the systemic effects.

amazing how easy it was to make unstable/chaotic systems, Robocop will be busy

Just adding one more link with a negative relationship between Political Unrest and Profits makes this much more interesting!

https://goo.gl/lL7EvY

To be realistic, more automation does not lead to "more profit", because of competition. But, this is really great tool to actually visualize these arguments.

yeah, visualizing lets us move the argument to a more focused context. we could examine and say, lets figure out if "more profit" really does lead to "more tax" money, etc

http://bit.ly/2obNrlR

Great. Automation also reduces prices, which makes unemployment more attractive. Eventually prices will be negative (automation produces more than anyone can consume).

Thus we have marketing and credit cards...

Mine is similar, but uses drugs: http://bit.ly/2nW8GbR

It's a Brave New World!

This thread is extremely entertaining!

I've got one surprisingly efficient, using fake jobs.

http://bit.ly/2n0CqDv

Makes you wonder...

We've been doing it for decades.

Yes, these jobs are part of the Tertiary Sector [1], but maybe we should open up a Quaternary Sector for them.

[1] https://en.wikipedia.org/wiki/Economic_sector

Need an extra + from robocop to "more profits". After all, you can lease the robocop to the government.

Or maybe just replace tax with robocop and really go to town.

This needs more work on the math side to be helpful. As it points out itself, the ecological example does not (in substance) correspond to the Lotka Volterra model.

Furthermore, there is no way for a species to go extinct from what I saw initially (perhaps there are options I'm missing?)

A lot of system-relationships are non-linear, or fractal, with time-ranges that are out of sync. Would be great to be able to model that -- i.e. introduce different "speeds" on how fast arrows can "travel" on the line, and to introduce functions that describe how x input arrows results in y output arrows, etc. It'd be great to be able to model how when a food source (or any input disappears), the burden shifts to the other sources etc -- now the arrows just go equally, sources/sinks can't disappear etc.

Yeah, I added another predator species, and I'm pretty sure the escalating cycles of feast and famine should have led to ecological collapse.

I think you can just draw longer arrows to reduce the speed. Or draw multiple to increase the factor.

For anyone interested in applying this stuff to real world issues, I highly recommend the +Acumen Systems Practice course [1]. Rob Ricigliano and Karen Grattan have put together a solid, pragmatic guide on how to understand, analyze, and engage complex systems. We've been working with Rob and Karen at Kumu [2] for years now and nobody else comes close when it comes to applied systems thinking.

[1]: http://plusacumen.org/courses/systems-practice/

[2]: https://kumu.io

What about the Coursera Process Mining class? I thought that was very similar, in concept, to this...

The technique used in LOOPY comes from system dynamics and is called causal-loop diagramming. Causal loop diagramming is a way to visually explore connections.

You can rigorously model systems using a stock-flow metaphor using desktop mathematical simulation tools like Vensim, Stella Pro, and Powersim. The system dynamics society has more information at: http://www.systemdynamics.org/what-is-s/

"Thinking In Systems: A Primer" was a wonderful introduction to the topic.

https://www.amazon.com/Thinking-Systems-Donella-H-Meadows/dp...

Yes, it's one of those very rare books that really made me think differently about things.

How much info does it give about how to practically predict and modify the behavior of realistic, complex systems? If I already understand that feedback cycles are important and hard to predict, how much would I benefit from this book?

My favorite kind of book is one that changes how you think about and see things, and this is one of those special books

Some relevant videos:

https://www.youtube.com/watch?v=eXdzKBWDraM

https://www.youtube.com/watch?v=HMmChiLZZHg

https://www.youtube.com/watch?v=0QtQqZ6Q5-o

https://www.youtube.com/watch?v=uYNlhjOZ7DU

https://www.youtube.com/watch?v=Pc3SWj-hjTE

https://www.youtube.com/watch?v=kz9wjJjmkmc

(the later ones are about Limits to Growth, co-written by the author of Thinking in Systems)

This is what I'm most excited about:

"LOOPY lets you have a conversation with simulations! You can go from thinking in systems, to talking in systems."

Thread: https://twitter.com/worrydream/status/808399253928218624

Love how the pencil/mouse gestures translate to adding nodes and links! That feature alone, even separate from the simulation aspect, makes me want a diagramming tool that does that for a wide range of symbols.

I'm sure I've used such systems in the past...I think even Flash used to have a similar feature. Anyway I love it too because it feels way faster than drag&drop.

Arrows travel with the same speed (pixels/s). So the shorter arrow, the stronger relation.

http://bit.ly/2mYqGle

You can also add multiple arrows right next to each other to make a stronger relation. If arrow is double the length but twice, it should be the same as if it's 1x length and just one arrow.

In some of the models people are posting, nodes were out of my window-view and I couldn't move over to them. Is this a bug or am I missing a scroll/zoom feature?

i think its missing pan / zoom. i had to move a chart from my laptop window to my big monitor window to view it all

You can zoom with the browser to see bigger models.

This is awesome. It reminds me of stuff I learned in a class called system dynamics ages ago: https://en.wikipedia.org/wiki/System_dynamics

This is awesome. I did a quick demo system to map Performance and UX to App Experience http://bit.ly/2nVWbNE My demo isn't great but this Loopy tool already has me thinking about the problem in a different way and attempting to map positive/negative inputs to a result in an interactive and shareable way.

I can't seem to scroll down to see your whole model on my particular laptop screen, is this a bug on the loopy site?

Same here, but you can zoom out with the browser and see the whole thing. OP is probably on a higher res monitor.

I played with the anxiety & depression example, but I didn't get it. Specifically, I don't understand how "doing things" leads to "feeling good" which leads to "depression".

Doing things has a positive (+) impact on feeling good, which has a negative (-) impact on depression. The sign there is important.

Doing things only has a positive impact if the outcome is at least positive.

Every minus sign inverts the direction of the change. Every plus sign leaves the direction of the change as it is. More doing things leads (+) to more feeling good, which leads (-) to less depression, which leads (-) to less de-motivation. But more de-motivation leads (+) to more depression, which leads (-) to less feeling good, which leads (-) to more depression.

A fair number of systems created here haven't grokked the behavior of (-), leading to models that don't actually behave as intended. I admit, it took me a few tries to notice, myself. The description given for relationships doesn't help, suggesting that (-) does nothing more than changing 🡩 to 🡫. Having three options, more (🡩), less (🡫), and flip (↕) would have been clearer and more versatile.

I can't find a good solution for the problem and that's a bit depressing... This feels a bit more accurate to me though: http://bit.ly/2nsIOXc

Beautiful and nice to use. After playing with the anxiety & depression model for a while, here's a thought: let me bring up magnitude/time plots for each of those circles. Density of arrows flying around the graph is not very readable. Adding proper plots (maybe hidden by default) could boost the exploratory capabilities of the tool without diminishing its ease of use.

That would be cool. Maybe you'd get a histogram by clicking on a node.

Imagine selecting two nodes and getting a Lissajous figure out of it...

Just found out that ncase.me has a lot of gems http://ncase.me/polygons/ and http://blog.ncase.me/we-become-what-we-behold-a-post-mortem/ are inspiring.

Nifty.

At a basic level, if applied to economic thinking, we get what preofessor Steve Keen has been working on with his Minsky software.

http://www.debtdeflation.com/blogs/minsky/

This is a brilliant idea. I really like the idea of bringing modelling into everyday thinking. I do not understand the idea of hiding the hard stuff. Why can't I have access to and change the code underlying the model too? Rich Hickey talked about code as a violin for the master or a one button instrument for the beginner. Perhaps we could have both?

I like the idea behind this, and I appreciate that it makes a complicated concept dirt-simple to use, but I think it's a little too simple. You can't diagram anything but a closed system, and most things are better modeled as open ones; the one addition of constant-up or constant-down nodes would make this much more useful.

You can make a node that feeds into itself, which is an okay substitute for an open system: http://bit.ly/2mwSlgX

Needs conditionals. Great concept though.

Politics and outside money: http://bit.ly/2mVfyod

Bipartisanship's effect on political power:http://bit.ly/2o9QieQ

> Bipartisanship's effect on political power

What's your thinking here? R and D are (presumably) zero-sum. Either with each other or with some 3rd/nth party. But there's a steady state in this system where R=D=0. Or are you modeling something other than actual seats held I guess?

I mean right now both R and D are extremely unpopular, probably in large part because of lack of bipartisanship (at least doesn't help). But R still hold a crap load of power.

Yeah, I guess my point was if you have bipartisanship it gives power to all factions. I thought it was a nice counterpoint to the system I shared with power swinging back and forth. No deep thought went into this.

Makes sense :)

Engaging with depression and anxiety as network effects rather than latent causes seems, anecdotally, to be productive. In my experience, the very networkiness of these things makes them hard to contend with emotionally, but treating the networkiness as a the culprit itself rather than a nearly-insurmountable barrier to contending with some real culprit has been promising. Scott Alexander on this way of modeling mental disorder here: https://slatestarcodex.com/2016/12/14/ssc-journal-club-menta...

If I am going to model systems that are finite state transducers instead of making me draw a circle why not have a click create the state and drag from one state to another to have an output.

When I am modeling a system its helpful to see what an increase or a decrease will propagate to the system but what is actually useful is some kind of chart that tells me cumulative results and also another chart possibly what is happening at this given second.

On your landing page your copy seems off. The first sentence implies that the system may or may not help.

I see this as an interesting toy but I don't see how this would be useful for modeling systems.

Sometimes I wish engineers in general would pay more attention to their "energy", and not the kind measured as the square of momentum but rather of the emotional kind. Is it really too much to ask that, in addition to requiring accuracy and precision, that you take time to consider the wisdom of your actions? I don't think so. If there is no joy in what you are saying (and I don't mean the joy of pointing out someone else's mistakes), why are you saying it?

Your comment is swimming in bad energy, but also in a kind of willful ignorance. What kind of ignorance? Of two kinds: first, you seemed to have missed the point of this project. It's goal is NOT to give you a way to model finite state transducers. It's goal is fundamentally pedagogical, and to give a wide-range of humans some experience with feedback loops, self-reinforcing behavior. This phenomena is abstract, and not widely appreciated. Second, you have misinterpreted the word "system" to mean "the specific production systems I am engineering right now", when in fact it was clearly meant more generally, as the demo systems show.

The damage you do with this sort of comment can be substantial. It is not easy to envision a thing and then make it and share it with the world, not least in part because bad energy comments, willfully ignorant and harsh, leave a mark. Creation is an act of vulnerability, and creators deserve our respect and all the positive energy we can muster.

A little nitpick - energy is not momentum squared: Momentum = mv and kinetic energy = 1/2 mv^2 So energy is proportional to mv^2 rather than (mv)^2

Hey, thanks for that. However, at the end of the day, mass is a constant. A constant squared is still a constant. To wit, for any given system, we can always pick units that make m=1.

> If there is no joy in what you are saying (and I don't mean the joy of pointing out someone else's mistakes), why are you saying it?

What about constructive criticism?

"Constructive criticism" has as it's prerequisite an understanding, respect and emphasis of the creator's goals, and an active deemphasis of your personal goals. This criticism had the opposite.

Using "personal goals" maybe be helpful feedback for the creator to understand things they had not considered from their own perspective. In this case, it might point to the goal of the project not being as obvious as it might be (I don't know if that's the case, just an example of what could possibly be constructive in this scenario).

The point of my criticism is not to project negative energy to the creator or taking joy in my comment but to provide the creator ways that they could improve the project. Under your line of thinking everyone should just go and project good energy and never offer ways for projects that they show to other people how that they can improve. I have posted my work on HN and the comments are very insightful for what I should improve or what I have done wrong.

If constructive criticism was your intent, it would have been constructive to make that intent explicit.

Especially in written communication, if one wants to provide constructive criticism (as opposed to criticism which is indifferent or just being critical to be critical), then it is important to signal that somehow. Possibly with words ;)

Of course not signaling that doesn't preclude the receiver of said criticism from using your critique constructively, but, imho, considering the minimal time needed to say something like "This is intended to be constructive criticism," and that doing so is not some sort of "happy-good-energy-no problems" use of language at all (your critique can still be as vigorous as desired), imho, there is no good reason not to and many good reasons (never knowing where the other person on the internet is coming from, improving one's own communication skills, increased explicitness, etc) to do so.

You offered a list of negativities that all ignore the author's intent, then capped them off with a statement that patronizingly declares that his project is so worthless as to not deserve existence. That is not constructive criticism. I cannot judge your intent - perhaps you meant well. At the very least I hope you meditate on the difference between destructive and constructive criticism, the key difference being respect of the author's goals, rather than your own.

his comment is effectively constructive criticism. stop marking him out to be such a terrible person for criticizing something that someone has authored. no work is immune to that, and creators don't deserve our "positive energy" if their creations, in our eyes, leave something to be desired. you don't get praise just because you created something, nor do you deserve it.

please don't defend thin-skinned creators, and also please don't speak for them when you don't know or understand their intensions.

> If there is no joy in what you are saying [...] why are you saying it?

In addition to my other comment, I think the "only say positive things to someone" attitude is dangerous.

I think it can be really damaging to someone if they grow up always only hearing positive things about themselves. It can lead them to be very egotistical and somewhat delusional -- they're great! they've always been told so! -- and unable to handle any form of criticism. And because they haven't received criticism they haven't been able to grow and develop as much as they could have, which can lead to incompetence.

This is great! I did quite a bit of work in system dynamics during my masters in social work -- one thing that has been missing from the world of computer simulation has been good open source tools and also tools that can start taking causal loop diagrams and turning them into simulations that can be played with. Excited to check this out, bravo to the dev.

I think that "systems literacy" is the literacy of the 21st century, and understanding feedback loops is a key part of that literacy (rather than just thinking in terms of linear cause and effect)

Would it be possible (useful?) to add a counter in each node to see the cumulative effect? I know that would be a slippery slope because ideally you'd need units and perhaps rate vs counter/gauge but it seems even a naive implementation of times fired would be instructive. I suppose since it's a graphical system the circles could expand/contract instead/as well.

This is a neat idea. I'd rather learn from simulation than, what is sometimes called, hard experience.

Is "systems" the academic field that studies real systems by simulating them?

Did UML come from this field. Do you know how UML is regarded nowadays?

Can someone educate me "What thinking in systems" mean?

https://smile.amazon.com/Thinking-Systems-Donella-H-Meadows/... is a great introduction and a relatively quick read

http://www.systemdynamics.org/what-is-s/

Basically, a while back, some people realized that all systems (which comprises just about everything, basically) share similar modelable attributes.

Who can solve the ticket horror show ? http://bit.ly/2n0OF3f

Looping with PhDs: http://bit.ly/2nVyhBX

I wish this had been around years ago.

Can someone draw a schematic for simulating Universal Basic Income?

what is being used to make animations? webgl?

It looks like it's based on a custom canvas implementation. The code is actually very written when you start poking around in the source.

Would be interesting to see the 'systems' for various start-ups. And compare those that were successful with those that were not.

Reply on news.ycombinator.com