Psychology and sociology experiments are unique among the sciences in that the test subjects who volunteer for them are generally capable of understanding the implications of the experiments' results. If the premise of the experiment requires deceiving the test subject in some way - as must be done sometimes, in order to simulate a situation for the test subject to act in - it may be that upon full disclosure of the experiment and its results, the test subject's reactions at the time may be cast in an entirely new light. This phenomenon is known as inflicted insight, and experiments which may cause it are strongly discouraged if not outright forbidden by ethics committees.
My question is - why?
What inflicted insight basically amounts to is exposing the less savoury or comfortable facets of a person's own psyche and forcing them to confront them. How could this be considered an undesirable result? The test subject believes something about himself or herself that is wrong, and the researcher is not only helping him or her to Know Thyself - in the best possible circumstance I can think of, I might add - but also preparing evidence that this problem is possible, for the scientific community, to better help other people who may have a similar reaction but were not fortunate enough to participate in the experiment personally.
Here's an example. The Milgram experiment is a famous one in which the unknowing test subject is asked to help with a simple negative feedback test for a 'slow learner'. Whenever the 'slow learner' - portrayed by an actor - gives an incorrect answer, the test subject is to press a button which, they are told, will deliver the learner a 15V electric shock that increases by 15V every time the button is pressed. The test continues until either (a) the test subject refuses to continue a total of five times (a list of four specific responses for the experimenter to use are prepared beforehand; after the experimenter has used all of them and the test subject still refuses, the test ends) or (b) the test subject uses the highest possible setting - 450V.
The hypothesized percentage of test subjects who would complete the test was 1.2% (with a range of 0% to 3%). While every test subject questioned the experiment at some point, 65% of them administered the final 450V shock rather than give up, and of the 35% who gave up none of them went to check the health of the learner or insisted that the experiments cease entirely with other test subjects. (They were informed of the true nature of the experiment afterwards during debriefing.)
Other variations on the experiment demonstrated that there was no significant difference behaviourally between male and female subjects, though female subjects reported higher levels of stress, and indeed there were only three factors that seemed to influence the outcome: compliance decreased when the subject and researcher communicated by phone (greater separation), when the experiment was conducted by an unknown institute rather than Yale (though the difference was not statistically appreciable), and when the subject interacted more directly with the victim. Experiments were also conducted in which the button was real and connected to a puppy, to eliminate the possibility that the subjects in earlier tests were subconsciously picking up that the button did nothing, with no change.
Such experiments would not be permitted today due to reasons of inflicted insight - taking the form in this case of "Here, mate, you just proved you would have electrocuted someone with a learning disorder because an authority figure told you to. And so did most of the other people who signed up to do this. Don't you think there might be something wrong with that?"
There is a coda to this story: a later survey conducted revealed that 84% of responding test subjects were 'glad' or 'very glad' to have participated, and 15% were neutral. Many in fact wrote letters expressing thanks.