Self-Identity Is a Schelling Point

Previously on The Scintillating But Ultimately Untrue Thought ("The Categories Were Made for Man to Make Predictions", "Reply on Adult Human Females"), we've considered at length the ways in which the self-identity criterion for gender (e.g., "Women are people who identify as women") fails to satisfy some of the basic desiderata for useful categories: the cognitive function of categories is to group similar things together so that our brains can make similar predictions about them under conditions of uncertainty. In order to make the case that it's useful to think and speak such that "identifying as" a gender is the same thing as being of that gender, one would need to show that those who identify as a gender form a natural cluster in configuration space—and not just a uselessly low-dimensional subspace thereof. ("Identifies as a woman" clusters with "prefers she/her pronouns", but if there's nothing else you can say about such people, then it's not clear why we care.)

Interestingly, a extension of this line of reasoning suggests an apparently novel argument in favor of the self-identity criterion—and which might go part of the way towards explaining many people's favorable attitudes towards the self-identity criterion, even if they've never formulated the argument explicitly. Let me explain.

(And please don't tell me you're surprised that I'm inventing novel arguments for the position I've spent the last twenty months of my life obsessively arguing against! Policy debates should not appear one-sided: it is by means of searching for and weighing all relevant arguments, that one computes the optimal policy, and even generally terrible positions will have some arguments supporting them. What did you take me for, some kind of partisan hack?!)

As, um, my favorite author on Less Wrong explains, another desideratum for intersubjectively useful categories is being easy for different people to coordinate on: in order to work together and think together, we don't just want to choose predictively-useful category boundaries, we also want to make the same choices.

The author gives the age of majority as an example. Presumably the right to vote should be based on relevant features of a person (in a word, "maturity"), not how many times the Earth has gone around the sun since they were born. But it wouldn't be practical for everyone to come to consensus on how to assess "maturity", whereas it is practical for everyone to come to consensus on how to subtract dates, so our shared socially-constructed category of "legal adulthood" ends up being defined in terms of a semi-arbitrary age cut-off, at the cost of mature 16-year-olds and immature 20-year-olds losing out on or gaining privileges that they should or shouldn't have (respectively).

When people need to coordinate on making the same arbitrary-on-the-merits choice, they tend to converge on an option that is (for whatever reason) unusually salient. This is the concept of a "Schelling point", after famed economist Thomas Schelling, who posed the question of where strangers should attempt to meet in New York, if they couldn't communicate to pick a rendezvous point in advance. The plurality answer turns out to be "noon at the information booth at Grand Central Station", not because of any properties that make Grand Central Station an objectively superior meeting place that you would pick even if you could communicate in advance, but just because its centrality makes it the focus of reasonable mutual expectations about what you and your partner are likely to do. Similarly, noon is salient as the midpoint of the day. There's no particular reason to meet at noon rather than 9 a.m. or 11 a.m. or 3 p.m., except that choosing 9 or 11 or 3 would seem to demand a particular reason that you expect your counterpart to be able to derive independently.

We usually expect the question of what sex (or "gender") a person is to have a canonical answer that everyone agrees on: it would be pretty confusing for bystanders if I thought Pat was a woman and said "Pat ... she" and you thought Pat was a man and said "Pat ... he."

For transgender people who consistently pass, this (ex hypothesi) isn't a problem. Unfortunately, in the absence of magical perfect sex-change technology, not all aspiring trans people pass consistently: the same person might be perceived as their developmental sex or their desired gender, depending on which observer you ask, how long the person has been on hormone replacement therapy, whether the observer knew the person before transition, the current lighting, or any number of other factors.

If, despite this, social reality continues to require the question to have a definite canonical answer, and we can't appeal to "passing" because that's too subjective and blurry, the natural Schelling point is, "Just ask the person what gender they are, and that's what they are." Even if we don't assume that people know themselves better than anyone else, people are still the focal point for reasonable mutual expectations about knowledge about themselves: if I claim to know Pat's gender better than she knows herself, and you claim to know Pat's gender better than he knows himself, then there's no more obvious way to break the symmetry except to defer the question to Pat.

(Notably, this is also the procedure you would use for non-trans people who just happen to be really-really androgynous: you're going to believe their answer to "Are you a woman or a man?" because if you could tell, then you wouldn't have asked.)

Schelling points are "sticky." If the set of possible choices is ordered, and it's possible to "move" from a currently-selected choice towards a "nearby" one, then the selected option may slide down a "slippery slope" until stopping at a Schelling point. Imagine the armies of two countries fighting over contested territory containing a river. The river is a Schelling point for the border between the two countries: unless one of the armies has a military advantage to push the battle line forward to the next Schelling point, we expect peace-treaty negotiations to settle on the river as the border. There's no particular reason that the border couldn't be drawn 2 kilometers north of the river, except that that would invite the question of, "Why 2 kilometers? Why not 1, or 3?"

The coordination problem of how to decide what "gender" a person is, can be seen as a particular case of the problem of how to decide what gender a person is in a particular context. The notion of the same person's "gender" being different in different contexts may seem strange, but again, in the absence of magical perfect sex-change technology, we might need it for some purposes: as far as the practice of medicine is concerned, for example, there's no getting around the fact that pregnant trans men are female. (Even if the doctors address the patient as "Mr.", "he", &c., they still need to draw on their mental models of the human female body to practice their craft, which presupposes a referent for the concept of "human female body.")

But here we have a slippery slope on what domains within Society should use developmental-sex categories or self-identity categories.

At one extreme, a "Sex is immutable and determined by the presence of a Y chromosome, no exceptions" regime is a stable Schelling point: if you have a lab that can do karyotypes, there would be no ambiguity on how to classify anyone with respect to the stated category system. (It would be cruel to trans people and people with complete androgen insensitivity syndrome, but it would be a Schelling point.)

At the other extreme, "Self-reported self-identity only, no exceptions" is a stable Schelling point: given the self-identity criterion of "Just ask the person what gender they are", there's no ambiguity about how to classify anyone. (This requires us to affirm the existence of "female penises, female prostates, female sperm, and female XY chromosomes", but it's a Schelling point.)

In contrast, any of a number of "compromise" systems, while potentially performing better on edge cases, suffer from ambiguity and are on that account less game-theoretically stable. It's a lot harder for Society to establish a specific convention of the form "Okay, you can have your pronouns, but you can't use your target-gender {bathroom, locker room, sports league, hospital ward, &c.} unless you {pass really well, get bottom surgery, have a gender recognition certificate, &c.}", not only because different factions will disagree on where to draw the line for each particular gendered privilege, but also because any line not drawn on a sufficiently sticky Schelling point will face constant attempts to push it up or down the slippery slope.