You're going to get yourself into hot water with that "trying" word in there. It's a mistake to look at this as a designed feature with a goal, when in fact what's going one is emergent, just doing whatever the heck it does, in concert with other features of the organism, in such a way that its chances of having successful offspring is increased.
As George Box put it, "All models are wrong, but some are useful".
Perhaps Aristotle and George would have been good friends?
I agree that using our agency-processing abilities can easily lead to wrong answers. But that's true of any of our abilities. We have a lot of hardware for visual processing, so we make graphs. Sometimes graphs mislead. But that doesn't mean we should stop using them. Instead, I think it means we should be aware of the limits of any particular cognitive mode and be energetic in cross-checking.
I think of it in terms of bias vs bias bias, and getting caught up in my own reflection; or, it's turtles all the way down (rationalism).