Crowd workers are distributed and decentralized. While decentralization isdesigned to utilize independent judgment to promote high-quality results, itparadoxically undercuts behaviors and institutions that are critical tohigh-quality work. Reputation is one central example: crowdsourcing systemsdepend on reputation scores from decentralized workers and requesters, butthese scores are notoriously inflated and uninformative. In this paper, we drawinspiration from historical worker guilds (e.g., in the silk trade) to designand implement crowd guilds: centralized groups of crowd workers whocollectively certify each other's quality through double-blind peer assessment.A two-week field experiment compared crowd guilds to a traditionaldecentralized crowd work model. Crowd guilds produced reputation signals morestrongly correlated with ground-truth worker quality than signals available oncurrent crowd working platforms, and more accurate than in the traditionalmodel.
展开▼