I quit porn three weeks ago and attempted to quit masturbation but failed. Subjectively I notice that I'm paying more attention to the women around me (and also having better orgasms when I do masturbate). My main reason for doing this was not so much that I found the research convincing as that the fact that people were even thinking about porn in this particular way helped me reorient my attitude towards porn from "it's harmless" to "it's a superstimulus, it may be causing a hedonic treadmill, and I should be wary of it in the same way that I'm now wary of sugar." (There's also a second reason which is personal.)

I like sixes_and_sevens' hypothesis. Here's another one: a smallish number of people really do have a serious porn addiction and really do benefit substantially from quitting cold turkey, but they're atypical. (I don't think I fall into this category, but I still think this is an interesting experiment to run.)

General comment: I think many people on LW have an implicit standard for adopting potential self-improvements that is way too high. When you're asking for conclusive scientific evidence, you're asking for something in the neighborhood of a 90% probability of success or higher. I think you should be willing to take probabilities of success in the neighborhood of 10% or lower in cases where the costs are sufficiently low. If you try out enough self-improvements, one of them may improve your life enough to have been worth all of the other failures (again, in cases where the costs are low). Plus, I think it's useful to make a habit out of changing your habits (think of it as simulated annealing on your life). Otherwise, you may just get better and better at arguing yourself out of changing anything.

In other words, I think people should be less risk-averse with respect to potential self-improvements. Anna thinks something like this is particularly likely to be a failure mode of people with a math background, where the demands for probability of correctness are much higher than in most of life.