Nicolas Kokkalis
Stanford University
Managing Personal Information with Private, Accountable Crowdsourcing

You need Flash player 8+ and JavaScript enabled to view this video.

Crowd-powered systems combine the power of human judgment and creativity with the speed and precision of computers. These systems can efficiently help people be more productive in ways that no single human assistant or computer program can do alone. Unfortunately, crowd-workers may not be able to help people manage personal information unless they have access to it and people may feel uncomfortable sharing such private data with online workers. I introduce privacy and accountability techniques for parsimoniously sharing private data with online workers, and provide experimental evidence that people can be more productive with the assistance of the crowd. These techniques develop crowdsourcing as a platform so that it is trustworthy and responsive enough to be integrated to personal information management. The work presented in this talk develops these ideas through two crowd-powered systems. The first, TaskGenies, is a task list that automatically breaks users' tasks down into actionable steps that they can complete one at a time. New tasks are decomposed through a crowdsourcing algorithm and tasks similar to existing tasks are selected using natural language processing. The second system, EmailValet, is a crowd-powered email client that introduces and applies the valet crowdsourcing technique for a user to share a limited subset of his inbox with an online assistant. The assistant reads the emails they can access and extracts tasks for the user. These systems point to a future where people are personally empowered by the crowd and crowdsourcing can be trusted with people's personal information.

Nicolas Kokkalis is a Stanford Ph.D. student graduating this year. His bio can be found here