It would be interesting to hear from zenman just what his intent was in posting that quote.
To me, it is not "love" vs. "science", which brings all sorts of false dichotomies into our practice, but a recognition that for all the rigor of our RCTs, we deal with individual patients in a "real world" that actually has to be controlled out of those RCTs. And so many of our patients are literally in the "low ground" socioeconomically...practicing here involves more swamp survival skills than academic accolades. The latest FDA indication for Cymbalta, or the newest enantiomer du jour, means little to the woman who is experiencing a cascade of bounced checks and bank fees because a state aid check bounced, and who must borrow to pay for the gas to drive to her clinic appointment.
I think the problem in psychiatry happens on a lot more basic level than just not being able to implement RCTs all the time. Even if RCTs could be theoretically applied in every treatment situation, we would still have a lot of ambiguity to deal with in psychiatry. After all, RCTs relate to treatment, which is the last step in a chain of clinical thinking. But before that, there are other steps. In diagnosing patients, things are often unclear, and in psychiatry we do not have many definitive tests. How often do I see patients who come in with "Bipolar d/o, Schizophrenia, and Schizoaffective d/o?" (It drives me crazy that actual psychiatrists let this stuff happen! It should be illegal with an increasing fine attached for each instance! So should discharges with unnecessary NOS diagnoses! Seriously, fines are the only way to stop egregious problems.)
But even on a more basic level than that, things can be ambiguous. For example, we might ask patients if they are "hearing voices." I've been reading Simm's Symptoms in the Mind, which is excellent for describing psychopathology. And I've had a course on the subject and many conversations with attendings and other residents, and I've asked many patients in excruciating detail what kinds of "voices" they're hearing. So you'd think I might be able to tell, just once! But NO! I cannot tell for sure if this is happening to anyone, although there certainly are clues, such as a person responding to internal stimuli, or acting on command auditory hallucinations, and of course, the things the patient tells me. But a Borderline patient will say they hear voices every day of their life, and I REALLY doubt that. A malingerer will say the same. A patient with tinnitus or a hearing abnormality might answer that they are hearing things too. So... it's subjective. As far as I know there is no test or algorithm. I suppose this type of subjectivity is part of the reason there are people walking around with psychotic diagnoses who in fact have something else going on. And this is way before the point of applying results of an RCT.
As frustrating as these situations in psychiatry are, I think there are parallels in other areas of medicine. For example, if you ask patients about chest pain, like what kind of pain it is and where it is radiating, that gets really mucky too. Someone once told me that those questions are actually not very reliable. The difference there is, they can follow up with tests to confirm/deny what the patient says. Whereas we don't have many tests. And just as a gross generalization--medical/surgical patients don't have the same rates of social problems that psych patients do. Nor do they face the same stigma. So psychiatry faces these unique ambiguous problems, and then a difficulty to implement whatever solutions have been shown to work.
Personally I see that as a challenge to address--not by "running away" from science or taking up romance novels or whatever, but by using whatever tools can be applied to the problem. For example, it is up to us to be more accurate in making correct diagnoses, and in adding as much clarity to situations as we can. Maybe one day someone will develop an algorithm (or at least a really great approach!) for dealing with the non-compliant patient. I mean, who knows?! And people should be advocating for social support for the mentally ill... which might improve compliance somewhat. I'm just saying, clearing the picture where we can will lead to less muck where it isn't needed, so we can apply research to the remaining scientific questions.
In any case, with a question like the one the OP asked, you are going to get different answers from different people...