Scott, known on LessWrong as Yvain, recently wrote a post complaining about an inaccurate rape statistic.
Arthur Chu, who is notable for winning money on Jeopardy recently, argued against Scott's stance that we should be honest in arguments in a comment thread on Jeff Kaufman's Facebook profile, which can be read here.
Scott just responded here, with a number of points relevant to the topic of rationalist communities.
I am interested in what LW thinks of this.
Obviously, at some point being polite in our arguments is silly. I'd be interested in people's opinions of how dire the real world consequences have to be before it's worthwhile debating dishonestly.
The problem with Yvain's reply is that he omits the main reason why lying is a bad idea. Yvain compares lying to violence. I don't think this is a good comparison. It's acceptable to respond to violence with violence. It's not a good idea to respond to lies with lies.
Eliezer touched on this issue in his post here where he pointed out that one problem with lying to support a cause is that you'd better be absolutely sure that all your beliefs about the cause and what to do for it are in fact correct. However, the problem is even worse, there is a vicious cycle here since a cause that frequently lies is much more likely to acquire incorrect beliefs.
Think about it this way: suppose you believe that your cause justifies lying, so you lie about it. Your lies attract people to your cause who believe those lies. They in turn make up further lies (that they think are justified based on the lies they believe to be true). And so no until your cause's belief system is full of falsehoods and anti-epistomology. Your cause may ultimately "win" in the sense that it's followers acquire power, but by that point said followers may no longer care about your original goal. Even if they do, they're likely to have so many false beliefs that what they do to accomplish it is likely to be counter-productive and probably have other unpleasant side effects.
Note that the above argument applies to lying but not to violence. Thus in some sense lying for your cause is in fact worse than committing violence for it.
One implication of this is that we can develop heuristics for how bad different lies. The basic idea is that lies that likely to spread (especially if their effectiveness depends on them spreading) are particularly bad. Especially if they're likely to spread within your movement (note lies used to increase support for your movement count here, since they'll bring in new recruits who believe them).
Note: that using these heuristics we can see that the classic example used to justify lying: "There are no Jews in my basement" is in fact much less bad then Yvain's example: "A man is more likely to be struck by lightning than be falsely accused of rape."