Gustavo Lacerda

It is consensus that sexual attraction is mostly heterophilic (genderwise) (i.e. people are "probably approximately straight") i.e. sexual networks are close to being bipartite by gender.
But SEXUAL adjacency patterns would suggest that sexual attraction is homophilic in most other ways (ethnicity, religion, class, age, body size, IQ, educational attainment, political opinion, "market value")
although this causal inference seems unjustified, since attraction is not the only cause of who one mates with (even after you forget about arranged marriages and such)


The following processes are all relevant:

* homophily (all kinds) causes similar people to be adjacent in the FRIENDS network
* proximity in the FRIENDS network causes memetic similarity (they begin to share a sub-cultural context)
* adjacency in the FRIENDS network causes liking ("familiarity breeds liking")
* proximity in the FRIENDS network enables courtship
* people will tend mate with those they like, as long as they are able to negotiate a successful courtship (I won't go into the complex dynamics of this)
* people tend to be more choosy about whom to add to their SEXUAL network than to their FRIENDS network (women more so than men). It seems likely that rather than simply requiring a higher standard of liking, this extra choosiness is due to extra criteria (e.g. age, IQ, etc.).

exploiting networks

Most dating sites attempt to match people by having them list the attributes they have and the attributes they like. This approach reminds me of good old-fashioned logic-based AI, and like the latter, it fails to capture the subtle and complex patterns (partly due to lacking data that users wouldn't enter, because they are themselves unaware of it; Facebook, OTOH, has such data).

Perhaps a wiser approach is to momentarily accept that attraction is a blackbox that we won't decode anytime soon, and find people who are similar to your type by exploring your SEXUAL network (similar in two senses: (1) attributionally similar, which is good for friendship (2) relationally-similar, which is especially good for sexual relationships (the SEXUAL relation is less symmetrical than the FRIENDS relation)).

Similarity-based dating: exploiting networks for attributional and relational similarity

Proximity in the FRIENDS network is a good way to find people who are friends material: they will tend to be similar to you in their interests and worldview. This is attributional similarity.

Now think of your more successful relationships, and look at their exes' exes (3 steps away from you). It's somewhat likely that they will have the same je-ne-sais-quoi as your ex that made you two right for each other in the first place (afterall, your alter-ego has already dated them both!). This is relational similarity.

If you're gay, there's the possibility of going 2 hops away (if you're straight, this will land you on someone who is the same sex as you). This could work well friendship-wise, but because 2 is an even number, it could be a bad way to find sexual partners! Even gay people show some heterophilic patterns in their sexual networks, I would bet.

Your alter-ego is probably (1) similar to you in attributional (friendship-relevant) traits (2) similar to you in relational traits (e.g. you probably share immune markers that make both of you attracted&attractive to the same type of person; this is a kind of heterophily, even though immune markers are not a binary trait.)

If the idea of being in a 4-cycle gives you the "yuck", then unless you've retired from the game, I would suggest you try to get over it.

---

^{Trackback: Peter Turney: "The Logic of Attributional and Relational Similarity"}