There is a popular illusion that confronting a computer with one’s ideas enforces rigor and discipline, thereby encouraging the researcher to reject or clarify fuzzy ideas. In the very narrow sense that the human must behave exactly like a machine in order to communicate with it this is true. But in a more useful sense, the effect is the opposite; it is all too easy to become immersed in the trivial details of working with a problem on the computer, rather than think through it rationally. The effort of making the computer understand is then mistaken for intellectual activity and creative problem solving.
Douglass B. Lee, Jr., “Requiem for Large Scale Models“
Journal of the American Institute of Planners
May 1973, Vol. 38, No. 3 (emphasis added)
7 Responses to quote of the week: on modeling
You are here: Home › quote of the week: on modeling
…and once “the effort of making the computer understand is then mistaken for intellectual activity”, the computer’s output (i.e. future predictions based on interventions, passage of time or both) is mistaken for the result of an intellectual decision on what the future will inevitably look like.
When doing modelling, my normal reaction to the output I see is “Do I like this”, then either “how do we make sure the future runs like the model”, or “how do we intervene to change that future?”
Modelling, at least at small scales (models that involve predicting macro-economic factors and people’s behaviour in 30+ years seem only to convince governments looking for easy answers) can be much better than predictions based on gut instinct. But after the model has predicted the future, then an intellectual discussion must happen on whether we like the results and what to do about it; perhaps going back to the model to input better decisions.
“”That’s where the truth lies, right down here in the gut. Do you know you have more nerve endings in your gut than you have in your head? You can look it up. I know some of you are going to say “I did look it up, and that’s not true.” That’s ’cause you looked it up in a book. Next time, look it up in your gut. I did. My gut tells me that’s how our nervous system works.””
Yeah. Here in Seattle-land we’ve had planners reading numbers out of computer models saying that the most riders-per-dollar comes from building trains along freeways stopping mostly at park-and-rides. You can’t argue with that — you don’t hate ridership or cost effectiveness, do you? — but if you look at the experience in cities that have done just this and ask, “Is this actually what we want to build?”, you might very well answer, “No.” It won’t actually create a city where you can make the trips you need to make daily on transit.
That’s a great point Al. At the end of the day, models are means to and end, but that end is still up to folks involved in the process to decide. Here in Denver, we’ve had a pretty aggressive (by US standards) rail expansion program, but I doubt anyone would argue it has substantially shifted (or will when it’s fully built out) anyone’s daily trip choices beyond commutes.
As someone who does alot of scenario analysis/GIS on the land use side, I’ve noticed alot of folks treat models in absolute terms: either they fundamentally trust them as deterministic or reject their accuracy/legitimacy. These are decision-support tools, not decision-making tools.
It’s also quite well-known that the best models are the simplest. In a particular sense: fewer parameters.
For instance, “intercity ridership is proportional to the sum of the squares of the populations of the two cities involved” is a useful model.
The Seattle transportation models have over a hundred parameters. The result of this is that they say whatever the planners running them want them to say.
“The result of this is that they say whatever the planners running them want them to say.”
As a layman observer, this seems to be the experience in Finland. The powers that be, usually highly placed public servants in conjunction with their favorite politicians, decide which projects they want to build, and then models are produced to show that those projects have a positive return on investment. When said powers don’t want to build something and need a superficially objective reason not to do it, a model is produced that shows the return to be slightly less than investment. This is no doubt easy to do by tweaking the numbers slightly.
E.g. the model for an orbital light rail line in Helsinki (currently a bus line) didn’t take land use benefits into account at all, which would otherwise be a major reason to build the thing in the first place. On the other hand, the reasoning for the heavy-rail metro extensions (which cost an order of magnitude more) relies entirely on land use arguments.
What I find most remarkable about this post is that the quote and the article it is from (“Requiem for Large Scale Models”) were published in 1973! If the author only knew how much further large-scale computer models would reach into our lives over the next 40 years. I’ll definitely want to check out the full article when I get the chance.