Your brain gets data, and has models of how all the data fits together. For example, if you look at someone you don't know, your brain will get the data "appearance" and plug it into the model "assorted stereotypes" and get out some expectations about who that stranger is.
What LSD does is perturb those models of how data fits together. If your models are in line with reality, you don't have much to worry about. After your trip, everything will settle back to about where it was before. If, however, your models are out of whack (e.g. you blame others for failures that are your own fault) then you run the risk of having some really really unpleasant realizations while you're looking at the data without your old model mucking things up.
I disagree, if only because I think some people diverge farther from reality when they take these drugs. I know some people who've gone from "slightly weird" to derelict insanity because of them.
If a person's already on a bad course, then psychedelics (under typical recreational use patterns) are going to throw this person farther out into space. The negative experience, in this case, isn't that person being ripped back into reality, but going even further away from it.
Could be, I'll admit I don't have much experience in this arena. I used the word perturb deliberately though, because it does not necessarily have to make you come closer to reality.
Your brain gets data, and has models of how all the data fits together. For example, if you look at someone you don't know, your brain will get the data "appearance" and plug it into the model "assorted stereotypes" and get out some expectations about who that stranger is.
What LSD does is perturb those models of how data fits together. If your models are in line with reality, you don't have much to worry about. After your trip, everything will settle back to about where it was before. If, however, your models are out of whack (e.g. you blame others for failures that are your own fault) then you run the risk of having some really really unpleasant realizations while you're looking at the data without your old model mucking things up.