I wanted to specifically flag this contribution by Kaimipono Wenger, however, which I think is particularly illuminating in explaining why these sorts of activities are a problem. One of the many things I learned from my basic statistics class was the concept of an "anchor". An anchor is a point of reference given to us when we are asked a quantitative question. So, for example, if I am asked "what's the average March temperature in San Francisco?" and then "do you think it is more or less than 70 degrees?" my answer is likely to be closer to 70 than if I been given no anchor at all. This is true even though the statement doesn't actually give me any true, relevant information.
Wenger first points out that the anchor effect still holds in seemingly absurd situations. In one study, the anchor was given via a wheel of numbers that was literally spun before the participant's eyes. In another, the anchor was something facially absurd: 558 degrees as SF's average temperature.
He then applies this logic to online intimidation even in situations where we might "rationally" tell the recipient that it is extremely unlikely they will actually be assaulted:
These and other similar tests show that the human mind is generally not capable of completely separating out and walling off information. Even if a person absolutely knows some piece of information to be false or to be irrelevant, it still enters into the total mix of information (to borrow a term from securities law) of her thought process.
This is why the rape and death threats are so insidious. They insert an extreme anchor value, intended to poison the total mix of information. Even if it is true that Kathy Sierra knows that these threats are bluster and hot air, they still enter the calculus for a host of very important questions (like “am I safe?”). And really, what could skew the calculus on those questions more than threats of rape and death? That is the reason why these attackers don’t simply say “I don’t like your argument.” That would be disagreement, but it would be in the normal expected range, and thus would not have the intense skewing effect of death threats. Death threats, on the other hand, are calculated to uniquely affect a person’s reaction, and behavioral studies tell us that they are likely to have some of that effect even if the target rationally knows that the threats are probably not serious. (The same may apply to other extreme anchoring speech; cf. Richard Delgado on racist speech.)
This of course raises the question of how to respond. There is a whole school of soft paternalism – Sunstein, Thaler, et al – who argue for a set of rules that subtly nudges people towards making better decisions. I would suggest that, in addition, hard paternalism may be required here. The damage done by the targeted extreme anchoring evident in Kathy Sierra death threats cannot simply be negated by saying "get over it" or "it's just bluster."
Fascinating article, fascinating research, fascinating analogy.