I hate cognitive biases. I read your comment right before I went to take the test. “Ha!” I thought to myself, “clearly members of Less Wrong wouldn’t be as effected. Why even bother mentioning it?” And then I clicked on the link while I thought about the singularity. “Hmm, 2100 is a decent year maybe it’ll be 20 years before that though...” And I filled in my race/education/sex. “Hmm maybe it would be after that though, due to...oh god, it’s the anchoring effect! Quick think of other numbers! 2090! 2110! Damnit. 1776! Wait that won’t work...”
And as I slowly worked my way down, by brain tried in vain to come up with alternate years. Until I finally reached the problem. “Is this really what I think, or am I just putting this answer because of that comment in the thread?” But it didn’t matter. The numbers were in the box, and I couldn’t convince myself to change them.
There it stood: 2100.
PS. Yvain, any chance you could look at the mean/median/mode/standard deviation of that problem before and after you changed the questions around? I’d be very interested in seeing how people were effected by anchoring.
Shouldn’t you ask when the respondent thinks the Singularity will occur before mentioning the year 2100, to avoid anchoring?
I hate cognitive biases. I read your comment right before I went to take the test. “Ha!” I thought to myself, “clearly members of Less Wrong wouldn’t be as effected. Why even bother mentioning it?” And then I clicked on the link while I thought about the singularity. “Hmm, 2100 is a decent year maybe it’ll be 20 years before that though...” And I filled in my race/education/sex. “Hmm maybe it would be after that though, due to...oh god, it’s the anchoring effect! Quick think of other numbers! 2090! 2110! Damnit. 1776! Wait that won’t work...”
And as I slowly worked my way down, by brain tried in vain to come up with alternate years. Until I finally reached the problem. “Is this really what I think, or am I just putting this answer because of that comment in the thread?” But it didn’t matter. The numbers were in the box, and I couldn’t convince myself to change them.
There it stood: 2100.
PS. Yvain, any chance you could look at the mean/median/mode/standard deviation of that problem before and after you changed the questions around? I’d be very interested in seeing how people were effected by anchoring.
Also possibly better to ask if before when for the same reason. And differentiate between blank = ‘it will not occur’ and ‘no opinion.’
If the survey is still going on, might want to remove your mention of the year 2100 as well, also to avoid anchoring.
It’s not.