This comment probably got downvoted for unnecessary ranting and sarcasm, even though the point it makes is a valid one: a published study has more credibility than a speculation on a forum.

For the sake of Nitpick, I'll first argue that I neither ever read Dilbert cartoons nor ever used such a thing as "pure reason" to my knowledge nor wrote any of this from within or benefiting-of-the-comfort-of my own home. However, since I'm just pointing this out, you should, if you persist in this course of argumentation, completely discard what I just said and assume that I did somehow.

The reason I did not present factual evidence, for my part, is that I considered it unnecessary on the prior that it be unlikely that someone who has read Eliezer's Core sequences (and reflected while doing so) would disagree, if only upon the notion that any prior in favor of "a lot of smart people have thought of this before us and yet we're still using it so it must be right" has already been shown in said sequences to be biased.

Notice that you've also completely ignored my main point and built a massive, chain-woven strawman painted black standing in the middle of the highway. The primary argument of my comment is that it is not the science which is entirely wrong, but the way the elements of the system fail to even acknowledge that there is something better they could be doing to select employees. Namely, employers being stupid. I back this up very weakly with anecdotal, statistically-insignificant and underpowered "evidence". Is there any more convenient a world you would wish for?

And here I was hoping that someone would rebuild my argument in stronger form before giving me reason to reconsider by showing that stronger argument wrong.

Right now, I have weak evidence (apparent lack of rational decision-making regarding employee selection on the part of employers) that the system is insane, yet strong evidence that it is at least not entirely sane in all situations. Conversely, there is no evidence suggesting to me that the system is "Sane", and every other variable that I suspect is correlated to this system's sanity shows indirect evidence towards insanity (examples in politics and religion come to mind most immediately, followed by various forms of warfare, systemic abuse and wilful neglect).

What's more, the system being sane is, in my opinion, only trivially relevant if it remains inefficient and sub-optimal due to lack of awareness of key variables that are, in hindsight, absolutely crucial and would be the first thing I go for. Naturally, the cost of learning this missing data is unknown at present, its deviation range being too large for me to even make a good educated guess.