I was interested to see what discussion this post would generate but I'm a little disappointed with the results. It looks like further evidence that instrumental rationality is hard and that the average lesswronger is not significantly better at it than the average person without a particular interest in rationality.

I'm going to throw out a bunch of suggestions for things that I think a rationalist should at least consider trying when approaching this specific problem as an exercise in instrumental rationality. I anticipate that people will immediately think of reasons why these ideas wouldn't work or why they wouldn't want to do them even if they did. Many of these will be legitimate criticisms but if you choose to comment along these lines please honestly ask yourself if these are ideas that you had already considered and rejected or whether your objections are in part confabulation.

One obvious reason for not trying any of these things is that the issue is just not that important to you and so doesn't justify the effort but if you feel that way ask yourself how you would approach the problem if it was that important to you. I haven't tried all these things myself. I rejected some as either too much effort for uncertain return on investment or in some cases had ethical qualms about them but I think they are the kind of things that anyone serious about instrumental rationality should have at least considered.

One thing that immediately jumped out at me as something of a hobbyist photographer was the casual remarks that people are 'not photogenic'. It seems to me that the word 'photogenic' should be like a red flag to a rationalist bull. It should immediately trigger a desire to unpack the meaning of the word and figure out what objective properties of reality it is describing. In this context the next response should be to figure out what elements that contribute to this concept are most amenable to conscious, directed efforts to fix.

What people generally seem to mean by 'not photogenic' is 'the pictures I've seen of this person do not seem to reflect the level of attractiveness that they possess in person'. Presumably people who are 'not photogenic' are not made of some different type of material that reacts differently to light than photogenic people. The problem must either be a lack of good quality photographs or an issue with uncomfortable body language when being photographed. Both of these are fixable given sufficient effort. I get the impression that at least some people in the thread didn't take the relatively low cost steps of reading OkCupid's advice on this issue or used the tool they provide for determining the picture that works best from the ones you have available.

OkCupid provides lots of data on OkTrends about what traits are considered attractive, broken down by gender and in other ways. With a little bit of research on this topic it is possible to make a list of areas where you could increase your attractiveness to the average person of the age, gender, etc. you are interested in attracting. Some of these are hard to fix (it is difficult for a man to make himself taller or a woman to make herself younger) but others can be improved with effort and are worthwhile goals in themselves (losing weight, increasing your salary). Figure out what the best 'bang for the buck' improvements appear to be for your particular situation and goals and expend effort on them.

A/B testing is a standard approach to optimizing online material. With a little effort it is possible to apply this to an online dating profile. At a bare minimum you can track any changes you make and record statistics on what improves your results and what makes them worse. If you wanted to get serious about this you could generate multiple profiles in different cities with similar demographics to your own and run parallel A/B tests rather than serial ones (this is one of those 'ethical qualms' approaches I mentioned). There are all kinds of shortcomings with the data collected in this way and with properly controlling variables but if you're not collecting any data of this kind you are not maximizing the information you extract from the data potentially available to you.

While the data that sites like OkCupid make available is helpful there are lots of interesting questions that it doesn't provide answers to. This being the Internet you could gather some of this data yourself. If you want to know what your competition looks like you could set up a fake profile for the kind of partner you wish to attract and see what kinds of messages they get (those damn ethical qualms again). This approach is potentially scalable to generate quite large amounts of data.

So if we're all good instrumental rationalists why are we not doing these kinds of things? Well for one, they involve effort. Quite a lot of effort in some cases. Instrumental rationality is hard. If we're not asking ourselves these kinds of questions though we're not doing a very good job of instrumental rationality. How can we improve?