I’ve followed this research closely. It has been run on thousands of subjects in carefully controlled multiple trials over a few years. The methodology is meticulous. The findings have been published in peer-reviewed journals. This Stanford Magazine article touches on it. I liked the illustration. The finding is simple — using the same morphing software you can buy online for twenty or thirty dollars, researchers morphed subjects’ faces into the photographs of political candidates. Without knowing that their own faces had been incorporated, potential voters who either weakly favored one candidate or had not yet made a decision decided to choose the candidate that subliminally resembled themselves — even though they were not told about the morphing until after they made their choices. Candidates can be made to look more caucasian, asian, etc. — depending, I presume, on the precinct:
…a Bailenson experiment done before the 2004 election in collaboration with Shanto Iyengar, director of Stanford’s Political Communication Lab, partially morphed the photos of undecided voters with those of either George Bush or John Kerry. Voters preferred the candidate they’d been morphed into—but could not consciously detect that the photo contained their own face.
I’ve followed this research closely. It has been run on thousands of subjects in carefully controlled multiple trials over a few years. The methodology is meticulous. The findings have been published in peer-reviewed journals. This Stanford Magazine article touches on it. I liked the illustration. The finding is simple — using the same morphing software [...]














Comments
@ 15:14
This BEGS the question how easy would this be to automate using Facebook or Myspace profiles, facial recognition software and say a quiz. People get an invitation to use a political quiz application on facebook, the application grabs your photos takes one tagged as including you morphs the face it finds into the politician the quiz is designed to bias you towards and you take the quiz…
Might not be quite feasible yet but it can’t be that far off.
@ 15:36
Hi, Howard!
Makes sense to me — does it seem surprising or controversial to you somehow?
I’ll bet familial similarity is similarly strong as similarity to the subject her/himself, and that familial similarity would be stronger than ethnic similarity. So I wonder how useful morphing per precinct/ethnicity would be.
It would be interesting to see results from familial morphs — sibling/same gender, sibling/opposite gender, mother, father, etc. And partner/spouse! That would be really interesting.
Pete
@ 17:31
Hmm. I’ll have to go back to the original articles, I suspect, but I wonder if they used a control face. That is, will I find the candidate more appealing if some random dude is mixed in with it.
If they haven’t checked this, of course faces that approach an eigenface are more likely to be judged attractive. The face that contains all faces is probably the most electable face of all.