Via Charles Arthur, ACA Wiki:

The authors also suggest that ascribing blame or community sanctions may be less effective than offering community members a way to “save face” “without having to admit that they deliberately violated the community’s norms.” They describe a system called stopit designed at MIT to address computer-based harassment. When users reported harassment, the system sent a message to the alleged harasser claiming that the alleged harasser’s account may have been compromised and urging them to change their password. Here is the rationale given by Gregory Jackson, the Director of Academic Computing at MIT in 1994:

…

recipients virtually never repeat the offending behavior. This is important: even though recipients concede no guilt, and receive no punishment, they stop. [this system has] drastically reduced the number of confrontational debates between us and perpetrators, while at the same time reducing the recurrence of misbehavior. When we accuse perpetrators directly, they often assert that their misbehavior was within their rights (which may well be true). They then repeat the misbehavior to make their point and challenge our authority. When we let them save face by pretending (if only to themselves) that they did not do what they did, they tend to become more responsible citizens with their pride intact.