Creditworthiness: Determined by the Crowd

Fascinating article in the Economist about using a simple Mechanical Turk task to judge the creditworthiness of individuals - just by looking at their photos. Quote:

"For his research, Dr Duarte chose a site called Prosper.com. His intention was twofold: to see if physiognomy-based prejudices about creditworthiness existed and, if they did, whether they were justified. To do so he used another peer-to-peer website to subcontract the job of assessing creditworthiness to a number of ordinary people. The site in question (which is run by Amazon) is called Mechanical Turk. 

The team recruited 25 Mechanical Turk workers and asked them to assess pictures of potential borrowers that had been posted on Prosper.com. In particular, they were asked to rate, on a scale of one to five, how trustworthy these people looked, and to estimate the percentage probability that each individual would repay a $100 loan. They were also asked to make several other assessments, such as the individual’s sex, race, age, attractiveness and obesity. The 25 results for each photograph were then averaged and analysed.

The researchers looked at 6,821 loan applications, 733 of which were successful. Their first finding was that the assessments of trustworthiness, and of likelihood to repay a loan, that were made by Mechanical Turk workers did indeed correlate with potential borrowers’ credit ratings based on their credit history. That continued to be so when the other variables, from beauty to race to obesity, were controlled for statistically."

Jacob and I were thinking about starting a category on the blog for "maybe this isn't such a good idea" ... This just feels unsavory. Although it's right in line with some of the great thinking in Gladwells' Blink and the studies on facial expressions done by Paul Ekman, for which I've got a tremendous respect and fascination.