Algorithm predicts Five Factor personality type better than Facebook friends

The Big Five has a lot of external validity (as do the regression estimates of Big Five, if you read the paper), so that's fine. If things like Conscientiousness really do predict income, longevity, etc, then even if they're a 'self-judgment', why is that a problem.

These self judgments also tend to be correlated with judgments made by friends, family, spouse, therapist, and with personality tests. In other words, even though there are differences between the groups, people are relatively self-aware.

I'd be interested to see here if friends in face-to-face contexts perform similarly to the Facebook context.

I don't think it's self judgement, they are asked to take a personality test which is a bunch of questions. The personality traits are literally defined as the way people answer the questions on the test.

It's not self-judgements, it's people expressing a like for something completely unrelated. Nobody "Like"d Nikki Minaj because they were hoping to show themselves as Outgoing and Active.

Right, the goal is to convince marketers that the digital shadow and the meat casting that shadow are one and the same.

In n' Out Burgers correlates with California, which correlates with NOT being from the Northeast, which correlates with being, relatively, calm and relaxed.

It's because you're looking at it from the wrong direction. The indicators are selected because people who are shy have a higher than average likelihood of liking "programmers". If you average out a bunch of likes and you know the general personality types of people who like those groups you end up with a pretty good general picture of someones personality.

I really want to see this run on one of those profiles that facebook marketing clickbots use.

I came to that conclusion when they told me "Sorry, it looks like you don’t have a sufficient number of Facebook Likes for us to generate an accurate enough prediction!"

That's a little off topic and I think you can probably find some relevant articles on the web but I remember reading about that before. As I remember, matching openness is pretty important for compatibility purposes, for neuroticism the important thing is that one or both scores are low, and for the others there are various advantages/disadvantages to matching up.