Hacker News Reader

Comment by seltzered_ | Hacker News Reader

This article is a lighthearted and doesn't talk about the underlying models behind simcity. Would recommend also reading:

https://logicmag.io/06-model-metropolis/ - which talks about simcity's inspiration from Jay Forrester's 'Urban Dynamics" and it's flaws.

Also, https://placesjournal.org/article/a-city-is-not-a-computer/ , which is something I've been meaning to read & contrast alongside Geoffrey West's works ( https://www.ted.com/talks/geoffrey_west_the_surprising_math_... ).

There's also a recent interview between an urban planner in portland (the citybeautiful youtube channel) and a developer at cities:skylines: https://www.youtube.com/watch?v=7c5pSCgbZ84

View on HN

The talks are nice.

The articles are very disappointing though.

The first one claims that Jay Forrester's urban model was wrong, because it produces right-wing/libertarian results. I may disagree with libertarians too, but that's not how you criticize a scientific model. How about some concrete stuff?

The second article is just bland. It asks a very good question: what should we optimize for when constructing a city? And then proceeds talking about different things none of which helps to answer the original question. The only takeaway: a city must have a library. A good thought, but not an eye-opener.

The argument isn't about results—it's that the model itself was built on ideological assumptions instead of anything empirical.

Well, I tried to find out what these ideological assumptions were, but:

There were no neighborhoods, no parks, no roads, no suburbs, and no racial or ethnic conflicts. - surely a simplification, but doesn't sound ideological.

Tax contributions of poor people are lower - well, again, this is probably a mathematical fact, rather than an ideological.

The ideological problem is assuming that all these factors don't matter—that taxes and zoning are the only factors that matter.

Some very fundamental examples of what's missing: Selective policing affects incarceration rates, neighborhood ethnic makeup affects city budgets which in turn affect land values and safety (Flint water crisis and other pollution) which in turn affect economic prospects and future tax income, what goes on in city A in terms of taxation and opportunities affects neighboring city B, it's assumed that traditional zoning policies generate the most land value which is incorrect in many cases, it ignores how public spending on infrastructure affects economic opportunities in the short an long term..

I am as far left as we go in France but that's a recurrent leftist criticism that I keep seeing and that I think is really misguided.

I do think that right-wing libertarians are wrong because they have a much simplified model of the world, that does not take into account all that you mention.

Scientists use simple models because they have technical constraints. Even more so for game designers who have to make a fun game while staying a bit realistic. They are biased, but not politically: they are biased technically, toward models that present a simplified model of the world that can be simulated reasonably.

It is easier to have a model of criminality where it is just a function of population density minus distance to police station than a multi-parameter model taking into account different social and ethnic makeup of the population.

But this is an interesting philosophical and actually scientific problem of the left: we do not have naively simplified models like the right has. Because free-market and capitalism are based on simple rules and an ideal but modelizable world.

This is also the reason why Adam Smith "The Wealth of Nations" had more success than his "Theory of Moral Sentiment" where he studies things like charity and non-selfish behaviors: the first one provides a mathematical model of how important variables change.

I would love to write a urban simulation that takes into account things like inclusiveness, corruption, inequalities but when it comes to actually code it, I don't know how to write a function of population growth based on that that would give results that make sense.

"Pop growth = 0.05% population * wealth factor" is much simpler. And when I think about it, I think that a leftist model is not a different model than a libertarian one, it is just a more complete one, I don't think a model that takes more humanitarian values into account can be as simple.

So please do not accuse every bias towards simplification of being a libertarian bias.

Sure, but what's the use in a simplified model that spits out a result that's wrong? Take the assumption that the "underemployed" pay little in taxes while taking money from the welfare system. Sure, on the surface. But that doesn't account for increased spending power as a result of money in their pocket (which isn't saved or put into a hedge fund, but actually spent immediately out of necessity) that goes back into the economy, nor does it account for upward generational mobility as a result of good health and affordable education, or a host of other factors that potentially turn the conclusion on its head.

The problem isn't with simplicity for simplicty's sake. My concern is that we're making public policy decisions based on models that are plain wrong

It is not wrong. It gives an approximation of reality.

With 2 or 3 differential equation, you can model the growth of a city and the densification of its center in a kinda realistic way. It gives you a baseline.

In such a model, population is a number. One way to look at it is to say that it has the assumption that everybody is equal, but another way to look at it is to say that we are averaging the population differences.

Such a model would fail at modelling some behaviors like the densification of some suburban island, would be totally oblivious to the creation of ghettos.

If the goal is to make a game, that can be enough. In some type of scientific or economic simulations they could be enough as well.

> The problem isn't with simplicity for simplicty's sake. My concern is that we're making public policy decisions based on models that are plain wrong

Models are, fortunately, not the only way to establish a fact-based policy. And far more complex models do exist taking more parameters into account.

> It is easier to have a model of criminality where it is just a function of population density minus distance to police station

I'm intrigued by this model in which the closer you are to a police station, the more crime there is. :D

If you like gaming-history you'll probably also appreciate:

https://www.filfre.net/tag/simcity/

Reply on news.ycombinator.com