Denying information can make the process better if the information is believed to have effect X, but it really has effect Y. For example, if the loan officer believes that all blacks are shiftless ne'er-do-wells, he might incorrectly downvote a mortgage from any African-American. Or, if he believes that all people who voted for George Bush are too stupid to pay their mortgage on time; that might also have a negative effect.
Essentially, there's a lot of unimportant, but easy-to-obtain information in the world. If someone uses it, it can easily make the loan predictions worse.
Now, of course, you're going to say that in a well-run corporation, the corporation would prevent that kind of thing from happening, in its own self-interest. True. But who says that they are all well-enough run to stamp that out? Do they really do the formalized statistical experiments necessary to decide whether each bit of information obtained from social media is really a useful predictor? Maybe, but maybe not.
And now, you're going to say, "But they have to be well run: the market forces them to be. Badly run companies go to the wall!". But that's not quite true. To survive, a corporation only needs to be 90% as good as its competitors. So, if the competitors aren't perfect, you don't need to be perfect either.
Therefore, I find it easy to imagine that less information can sometimes (even often) be better.