Priors are all about Priorities

When you go upstairs in an unfamiliar house to look for your friend's cell phone, there's a meta-step that goes along with your casting up of a prior: a cost-benefit analysis. Do you predict well enough that you can ignore most of the data and get away with very little tweaking—essentially look through a blurry, inaccurate lens and not care? Yeah, classic 1920's tread pattern, as you would expect in that house. Don't need much more detail than that. See the door on the right. Know how to manage the pivot in my sleep. Oooops! Who would have thought they would have painted their office walls crimson? … and so on. If you are in a hurry, not hyper-vigilant, don't care about architecture or interior design, aren't very nosy and are actually preoccupied with what is for lunch, you are going to take in the least information you need in order to accomplish the task.

Meanwhile, you are busily making the necessary adjustments to your scenarios as necessary. Maybe some of your assumptions were wrong, but don't matter. The wrong assumptions that matter—the tread is actually deeper—get a meep-meep-meep but the wrong assumptions that don't matter are ignored. If your system is calibrated about right.