Using Prosper and Mechanical Turk to figure out if people who are shifty look shifty

http://lsvp.wordpress.com/2009/03/19/using-prosper-and-mechanical-turk-to-figure-out-if-people-who-are-shifty-look-shifty/

by jeremyliew • 17 years ago

19 5 17 years ago

lsvp.wordpress.com

tdoggette• 17 years ago

Okay, so 'shiftiness' is independent of appearance as measured other ways, but what is it exactly? The next step is quantifying what cues people pick up on that identify trustworthiness, and that's what I'm really interested in.

ktharavaad• 17 years ago

An excellent way to figure this out would be to apply some sort of machine learning technique. Perhaps an Active-Appearance-Model or some other statistical measure to mathematically represent the shape and texture of an "untrustworthy" face as a lower dimensional vector. Then train classifier on this data which can discriminate between "trustworthy" and "untrustworthiness". Of course, there are going to be ethical concerns, but this is definitely one interesting way of doing it.

tkt• 17 years ago

This sounds a bit like Bruce Schneier's comments on "hinky" http://books.google.com/books?id=wuNImmQufGsC&printsec=f...

eli• 17 years ago

This blog post adds virtually nothing over the original linked article: http://www.economist.com/science/displaystory.cfm?story_id=1...

Otherwise, interesting.

alexandros• 17 years ago

Might this be a self-fulfilling prophecy? If the 'shifty' group were offered higher interest rate, then this increases the probability that they will bail on their loans, no?

Using Prosper and Mechanical Turk to figure out if people who are shifty look shifty

Related Stories