(no subject)

[randomness]

"You should change your model to fit reality rather than casting out data points until your model works."

You'd think this would be a fairly simple concept, but you'd be surprised how often and in how many ways people violate something this basic.

I spent a lot of my life with some exceptionally broken mental models: about money and work on the one hand, and about relationships on the other*. It wasn't until I started adjusting them that I started having more success in either field.

We can't get along without models of reality. Reality is just too complex for our limited brains, and models are a useful construct to allow our brains to make some kind of decision given complex reality. (Jonah Lehrer talked yesterday on Fresh Air about how people are overwhelmed by excessive complexity when trying to make decisions.)

My mental models still have a lot of issues, and I do have some concern that I may run into the problem Taleb keeps talking about, where your model is good enough in the current circumstances but may somehow be prone to abrupt and catastrophic failure when conditions you have assumed not to ever be worth worrying about turn out to be much more common than you expect. Then they blow up. (The model, your relationship, the financial markets, whatever.)

*I'll be glad to discuss these, of course, but this post was just about adjusting models to fit reality rather than rewriting reality to fit an existing model, as opposed to what the model itself is.

Comments

[r-ness.livejournal.com]

The way I see it, you're extending your model, which to me is a way of adjusting the model. You had a model which you thought was universal but now realize isn't. The realization that your mental model is not universal is in itself an adjustment to your model, even if you haven't yet got enough datapoints in the new situations to build out the model so it can be helpfully predictive.

That building out you're doing now by testing your model in the new contexts.

At least, that's how I see it.

My sympathies, btw. It's hard when one's interacting in a cultural context one has no experience with.

[redbird]

That sounds like what jab2 is doing, yes.

It's a bit like me noting that my relationship model works for a certain kind of person, but that "how we do this" is not only demonstrably not how everyone does this, there are a lot of people it probably wouldn't work for, and that's okay. Not, necessarily, that every way of doing relationships is right, but that there are quite a few that are, and more power to those who have found one of those that works for them and those they love.

[r-ness.livejournal.com]

It sounds like you're using "relationship model" to mean, "how I do relationships", where I mean "model of relationship" to mean "how do I think they work". Each informs the other, but I think of them as significantly different.

I agree with you aside from that, however. :)

[digitalemur.livejournal.com]

I have collected a certain amount of information about "how I think relationships work" in areas where I'm not so eager to incorporate them into "how I do relationships." This may be either because they were once part of "how I do relationships" and I've decided I'm not so comfortable with some of those things anymore, or because I've always watched them in other people and am pretty sure I don't want to incorporate them. All the same, while there are some things I really don't want to incorporate into "how I do relationships," I do like to understand how people work and why they might choose something that I know really isn't such a great thing for me.

[redbird]

That is a useful distinction, yes. Some of the ways that relationships work, I can only model as a black box (if input is X these people will produce output Y) without having any understanding of Y. Others, I can look at and think "Okay, weird axiom, but from there I can see how they get to that result/activity."

And that paragraph is using "work" to mean something that is more or less functional, as distinct from "if they do that again, they'll have another messy breakup," which is also part of a model.