Signals vs. causes
It turns out that people who use Firefox are more likely to engage in certain online activities than those that use IE.
And it turns out that people who eat before bed are believed to gain more weight than those that don't.
Perhaps using Firefox makes you a different sort of surfer, or the timing of the calories has something to do with your metabilism.
More likely: The sort of person who takes the time to install a new browser is precisely the kind of person willing to use a new web service. The kind of person who makes a habit out of eating when bored (just before bed) might very well be the kind of person that has to wrestle with weight.
We see the same thing in outbound marketing. Spammers in Nigeria continue to use poorly written, ridiculous pitches. Not because they cause people to give up their senses and send tens of thousands of dollars, but because the kind of person that falls for something so dumb is probably the kind of person who is also going to be easily scammed.
TED often attracts interesting people, but going to TED (love this hashtag) doesn't make you interesting.
People who order wine with dinner might be bigger tippers, but persuading someone to order a bottle probably won't change the way he tips.
A fever might be the symptom of a disease, but artificially lowering the fever (ice bath, anyone?) isn't going to do anything at all to change the illness.
Before changing the signal and thus assuming that this will change the outlook, it probably makes sense to understand what will change the causes of someone's perception and habits, and use the signal as a way of figuring out who needs to be taught.