Human-Centered Research at Crowd Scale Hive: Decentralized, Collaborative Design Through Network Rotation

PI: Michael Bernstein


Decentralized, collaborative design (e.g., OpenIDEO) taps into online volunteers to gather designs that are more varied than a single design team could produce. A core assumption is that people will draw on each others’ submitted ideas to increase their collective creativity. However, reality is starker: mere exposure to others’ ideas does not cause meaningful engagement with the ideas, resulting in mountains of mundane, repetitive submissions. In this proposal, we draw on organizational science demonstrating that external ideas exert influence not through their presence, but through interaction with the people who bring them. We propose a system that facilitates intermixing of ideas in a large crowd of contributors by gradually changing team membership to expose teams to new members and new perspectives, a process we call network rotation. Network rotation requires computing who should move where, and when, to draw together new perspectives: a combinatorial problem with exponentially many alternatives. We propose an efficient solution by computing the maximum cycle cover over a collaboration graph. A pilot field deployment of our system targeting an OpenIDEO challenge — reducing stigma around disabilities — generated ideas rated 1.8 standard deviations higher than the top ideas submitted to the actual competition.