Stanford Research Project Uses Crowdsourcing to Organize Inboxes

Handing over your car keys to a complete stranger is an accepted risk for the benefit of valet parking. But what about handing over access to your inbox for the benefit of increased productivity?

Researchers at Stanford University are finding that people could be willing to do just that—with the right security in place.

EmailValet, a graduate research project at Stanford, finds remote assistants through the crowdsourcing-for-hire Web site oDesk, then allows them to read a user’s messages and create a to-do list from the information they’ve read. Like some valet keys that allow parking attendants to open car doors and start the engine but prevent them from getting into the glove compartment or the trunk, EmailValet lets users select what kinds of e-mails their assistants can read.

The idea for the project came from Nicolas Kokkalis, a computer-engineering Ph.D. student at Stanford, who turned to online crowdsourced contractors to help sort through his messy inbox after starting his own company, in 2009.

“I found people who could help,” he said. “But how could you trust an online assistant from someplace far away?”

So Mr. Kokkalis, working with Stanford’s Human-Computer Interaction group, created an application that would solve those trust problems. Before a test run of the project, the group surveyed nearly 600 people to find out what attitudes existed toward handing over complete access to inboxes.

Only 4 percent of the respondents said they would be willing to hand over that much access, and half said they were uncomfortable with the idea of allowing a stranger to comb through their messages, said Michael S. Bernstein, an assistant professor of computer science at Standford and an adviser for the project.

But, Mr. Bernstein said, when the group conducted a study with 28 testers using EmailValet, there seemed to be a change of opinion. By the end of the study, only three people said they would not use the system.

“The most surprising thing to us was the large distance between what people said they would do and what they actually did,” Mr. Bernstein said. “It just reminded us that privacy is a dynamic practice, really. It tends not to be based on ‘Do I share this or don’t I?’ but more on ‘What do I get in return?’”

And the price of the return would also be relatively cheap, Mr. Bernstein said. The researchers found that one assistant could juggle as many as 36 inboxes at one time, he said, meaning the cost would come to just $2 a day for each user.

Another adviser for the project, Scott R. Klemmer, an associate professor, said the sense of trust came not only from the ability to decide what kinds of e-mails assistants could read but also the relationship that arises from interacting with an assistant more directly.

To Mr. Klemmer, this shows that “by providing privacy and accountability, we can enable a richer class of tasks to benefit from crowdsourcing.”

Mr. Kokkalis said that the research could point to a new way of looking at crowdsourcing.

“Crowdsourcing initially started as a field where we mistakenly viewed it as underpaid workers that do things for two cents,” he said. “I think that impression is wrong. We can definitely see a future where crowdsourcing is highly paid and a job with a high reputation.”

Return to Top