“Every model is wrong, but some are useful”

heartofaquamarine:

The title of this post comes from the statistician George Box, and discusses how the mathematical models we use to analyse and predict the world cannot, by their nature as an understandable simplification, provide an exact model of the world. That is to say, they are “wrong”. However, the errors can be small enough that we can use the models to make predictions that hold up in real life, such as how much weight a bridge can carry. That’s where the “useful” part comes in. A good example of this is the ideal gas law. No actual gas follows this law, but under certain conditions, their behaviour is close enough that it doesn’t matter.

An important thing to understand is you can have multiple models to describe the same thing. There are a vast array of different climate models, for instance, and while they broadly agree with each other there are important differences in the details. Some of these are due to limitations on computing power and time available, some are focused on certain aspects such as aerosol production, and some are simply using a new method of modelling the world. When we are dealing with subjects as complex as human sexuality or the climate, we need to simplify them somewhat to allow us to grasp, if not the whole concept, as much of it as we can. All of these models will lose some important details, hence why they are all wrong, but some retain enough of the information that we can use them to understand the world.

Quite a few other statisticians have issues with this aphorism, but that’s not what I want to talk about here. I want to talk about gender and sexuality, and how we think about them. A common comment, when we look back at history and retroactively apply certain traits and sexualities to them, is that they wouldn’t have chosen those terms because their internal models are entirely different. Parts of Greece and Rome, for instance, are said to have defined the roles in male-male sexual encounters in terms of who is being penetrated and who is doing the penetration. Nowadays we would define both partners as homosexual, but they used a different model to understand their interactions than we do now. Different people at different times, different places and in different circumstances will use different models Someone using a different model of sexuality and gender to you is not inherently bad, or a judgement on you. I say not inherently because there are absolutely models that fail that include various points. I’m going to use asexuality here as an example because that’s my own experiences, but I think the general points could be modified slightly and applied to most sexualities.

Incomplete models, that is, ones that cannot handle the idea of someone not being attracted to anyone, fall into two catagories. The first is the “innocently” incomplete model; it is not that the person disbelieves in asexuality, but rather the idea that there is a gap in their empirical model of reality just hasn’t occurred to them. Contrast this to what we might call the “active” incompleteness model, where the model actively denies the existence of asexuality. I’ve encountered both of these kinds of incomplete model, as I suspect a lot of people have, and while the models are superficially similar how they react to the addition of an asexual person, or at least a person who claims they are asexual like myself. I put the “claims to be” part in there not as a way of throwing suspicion onto myself, but rather to acknowledge that people will claim I am lying or mistaken, that it is not possible to not be attracted to someone (you can insert “attracted to your own gender”, “attracted to more than one gender” or any other phrase you like there). Both models are wrong in the same way but the second model has a built in mechanism to avoid expanding the model to take into account an aspect of reality it had been lacking.

The other thing about models is that, once you get beyond a certain point, adding more detail doesn’t help unless you are using a model for a particular purpose that takes that detail into account. The various terms and models of sexuality that we classify as being on the asexual and/or aromantic spectrums are important if you are in a situation that requires the distinction to be made. Usually this means you are, or are interacting with, someone under that umbrella or wish to compare the model to your own experiences to see if you might be under that umbrella yourself. For most people most of the time, adding this detail just adds more mental computing power, and it is enough to acknowledge that, yes, you might need to increase the resolution on your model when dealing with that particular area, just like a meteorologist looking at one particular region will increase that resolution, but for now your model is wrong, but useful in that it acknowledges the existence of the asexual spectrum.

A really good example of this is the romantic/sexual attraction divide. This is a really important distinction in the ace community, between romantic asexuals and aromantic asexuals, and it is also useful for people outside of it (for example, there are bisexuals who are only romantically attracted to one gender), but is useless if your romantic and your sexual attractions line up. It can even muddy the waters by introducing an alleyway that might not lead you to a helpful option. That being said, the mere fact that such a model is not useful to everyone does not make it useless, merely that we need to shift our discussions about such models from “this is how the world works” to “here are mental tools to help understand the world and your place in it, and here are some possible limitations to them”.

I think gender is even more complicated to model than sexuality just because frankly I think the boundaries we have set on our concept of gender are incredibly fuzzy. The less well defined something is, the more difficult it is to model accurately, and we are really bad at defining gender. I think the three main models are as follows: biological gender, which states that someone’s gender is predetermined by their biology, usually chromosomes or genitals, social gender, in which gender is defined by the range of social expectations allowed with in it, and individual gender, in which gender is a property of each individual, that must be defined by each individual and social gender dynamics grow out of these definitions and their statistical distribution. Notably I have listed this in increasing order of immediate complexity. Biological gender claims to be less an explanatory model and more an empirical observation; these are the two types of people. I’m not fond of it for a couple of reasons. Firstly it doesn’t actually position the information in a new or useful way; it is merely a description of biological factors, and the distribution curves for what it tries to predict are large enough and include enough other factors that breaking it down this way doesn’t add anything. Secondly, it fails to account for intersex and non-binary people. The two ways it handles these are either to ignore them as edge cases, which works up until you need to include them as part of the model, or to outright deny their existence, which again, reduces the model’s effectiveness at describing reality. To make matters worse for this mode, it isn’t actually a particularly useful model day to day; when choosing what pronouns to apply to someone (a process that in itself is a conceptual model; there are languages with only one, non-gendered third person pronoun), you don’t ask for a karyotype test result or compared genitalia, but rather make the assumption from visible signifiers which don’t inherently match the biological sex.

The second one, social gender can be useful in examining social mechanisms and constructs, but in doing so creates a lot more edge cases. Now it isn’t just intersex and non-binary people that must be accounted for, but gender non-confirming people who, by definition, are edge cases in the model. The social model of gender becomes particularly dangerous, rather than simply incomplete to an alarming degree, when it becomes prescriptive instead of descriptive. To paraphrase a certain first century carpenter “The model was made for the people, the people were not made for the model”.

The final model, the individual model, essentially simply increases the resolution as high as it can go, and relies on people’s internal definition of gender. Full disclosure; I am agendered, so I don’t actually understand what it means to have a particular gender, but some people I know have very clear internal definitions. Some of these are based on physical sex or social norms; I think there’s quite a few cisgendered people who lack an internal definition like myself, but don’t have a reason to use a definition other than the biological one. We could call this being cis by default. Other definitions of gender are highly idiosyncratic, and to be honest I think this is inevitable just because, like I said above, the initial parameters that we are trying to model are really badly defined. What do we mean by gender? What weighting do we put onto different aspects of it? And that’s before we get into the question of working out the mechanisms that give rise to it; untangling biological and social influences is difficult enough given we can’t isolate and raise babies in a controlled lab because ethics is a thing and is important. As such, I think the best way to proceed is to find your own model, but acknowledge that it is not universal and accept other people’s own models of their gender. It’s unsatisfying, but at this point, what else can we do?