Mathematics of talent

Google has developed new on-line technqiues for locating talent, including a web-based survey and new alogorithims to identity people who fit both the Google culture and specific kinds of jobs. Not surprisingly, the company has quickly learned that academic performance is not necessarily a good indicator of on-the-job performance, according to this New York Times' story.

"Desperate to hire more engineers and sales representatives to staff its rapidly growing search and advertising business, Google — in typical eccentric fashion — has created an automated way to search for talent among the more than 100,000 job applications it receives each month. It is starting to ask job applicants to fill out an elaborate online survey that explores their attitudes, behavior, personality and biographical details going back to high school. The questions range from the age when applicants first got excited about computers to whether they have ever tutored or ever established a nonprofit organization. The answers are fed into a series of formulas created by Google’s mathematicians that calculate a score — from zero to 100 — meant to predict how well a person will fit into its chaotic and competitive culture."

"Unfortunately, most of the academic research suggests that the factors Google has put the most weight on — grades and interviews — are not an especially reliable way of hiring good people... Last summer, Google asked every employee who had been working at the company for at least five months to fill out a 300-question survey. Some questions were factual: What programming languages are you familiar with? What Internet mailing lists do you subscribe to? Some looked for behavior: Is your work space messy or neat? And some looked at personality: Are you an extrovert or an introvert? And some fell into no traditional category in the human resources world: What magazines do you subscribe to? What pets do you have?... "

"One score was what the company called “organizational citizenship,” said Todd Carlisle, an analyst with a doctorate in organizational psychology, who designed the survey. That is, “things you do that aren’t technically part of your job but make Google a better place to work,” Dr. Carlisle said, such as helping interview job candidates. When all this was completed, Dr. Carlisle set about analyzing the two million data points the survey collected. Among the first results was confirmation that Google’s obsession with academic performance was not always correlated with success at the company. “Sometimes too much schooling will be a detriment to you in your job,” Dr. Carlisle said, adding that not all of the more than 600 people with doctorates at Google are equally well suited to their current assignments. Indeed, there is some resistance even at Google to the idea that a machine can pick talent better than a human. “It’s like telling someone that you have the perfect data about who they should marry,” Dr. Carlisle said."

 

Read the whole thing here.

Comments

[Alexandre Borovik]

At least some questions are obvious proxies for (apparently illegal) questions about age, gender, family situation:
* What Internet mailing lists do you subscribe to? . . .
* What magazines do you subscribe to?
* What pets do you have?

Give someone's answers to 300 questions like that to any decent sociologists, and he/she will tell the sex, sexual orientation, marital status, number and ages of children.

[Timtian]

Human beings are animals of signal. Otherwise the world will be too complicated for us. The problem of google is that it can’t find one or more reliable signals of the people it wants. A college diploma or an occupational certificate works for most traditional industries, but an innovative firm like google needs to select talent with an inclination of innovation and initiation, which can’t be measured effectively by the exams.
I don’t believe the machine can do a better job than HR interviewers because the formulas used for selection are also written under some assumptions which may be false. But the machine can reduce the pool of candidates and then the workload of HR component.