The paper introduces flash teams, a framework for dynamically assembling, managing, and scaling paid expert crowds to accomplish complex, interdependent tasks that are beyond the reach of traditional micro‑task crowdsourcing. Flash teams are expressed as sequences of modular blocks—each with typed inputs, outputs, instructions, and a directly‑responsible individual (DRI)—that can be linked into workflows resembling lightweight organizational structures. The authors present Foundry, an end‑user authoring and runtime platform that lets users compose blocks, recruit experts from oDesk, and automatically handle handoffs, notifications, and timeline management. Computational mechanisms such as automated path‑search (using STRIPS planning), elastic growth (replication and specialization), and pipelining of intermediate results enable on‑demand creation, scaling, and acceleration of teams. The paper showcases four families of flash teams—napkin‑sketch design, animation, education, and a composite online‑course platform—detailing block configurations, timelines, costs, and outcomes. A controlled field experiment comparing flash teams to self‑managed teams (same experts, same task) shows flash teams require roughly half the total work hours, exhibit lower variance, and achieve comparable quality with less coordination overhead. The discussion reflects on the broader vision of elastic, on‑demand expert organizations, acknowledges limitations (coordination challenges, recruitment overhead, need for trusted expert pools), and outlines future work on broader task domains, manager‑led teams, and richer runtime support. The conclusion argues that computationally‑guided expert crowds can extend the reach of paid crowd work from redundant judgments to coordinated, high‑skill collaborations.
Topics
Flash Teams: Modular, Computable Team Structures for Expert Crowds
Definition of flash teams as linked sequences of modular blocks with typed inputs/outputs, DRIs, and lightweight coordination structures.