Reputation-Based Mechanisms for Reliable Crowdsourcing Computation

摘要: We consider an Internet-based Master-Worker framework, for machine-oriented computing tasks (i.e. SETI@home) or human intelligence (i.e. Amazon’s Mechanical Turk). In this framework a master sends tasks to unreliable workers, and the workers execute report back result. model such computations using evolutionary dynamics three type of workers: altruistic, malicious rational. Altruistic workers always return correct result, always incorrect rational (selfish) workers decide whether be truthful depending on what increases their benefit. The goal is reaching eventual correctness, that is, stable state system in which it obtains results. To respect, we propose mechanism uses reinforcement learning induce behavior workers; coping with malice leveraging reputation schemes. analyze our as Markov chain we give provable guarantees under can ensured. Simulation results, obtained using parameter values similar the values observed real systems, reveal interesting trade-offs between various metrics parameters, cost, time convergence behavior, tolerance cheaters metric employed.