Mitigating Herding in Hierarchical Crowdsourcing Networks.

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Science

Background:

• Hierarchical crowdsourcing networks (HCNs) facilitate social mobilization but are prone to herding behaviors.
• Herding leads to reputable workers being overloaded while others remain idle, a problem not addressed by existing control mechanisms.
• The complex dynamics of resource allocation in HCNs necessitate novel approaches to manage worker participation.

Purpose of the Study:

• To investigate herding dynamics specifically within Hierarchical Crowdsourcing Networks (HCNs).
• To propose a novel decision support approach, RTS-P, to mitigate herding and enhance collective productivity.
• To achieve superlinear time-averaged collective productivity in HCNs through optimized task allocation and pricing.

Main Methods:

• Developed a Lyapunov optimization-based decision support approach named RTS-P.
• Incorporated worker reputation, workload, eagerness, and trust relationships into the RTS-P model.
• Utilized high-resolution simulations to evaluate RTS-P's effectiveness compared to existing methods.

Main Results:

• RTS-P systematically mitigates herding behaviors in HCNs.
• The approach facilitates joint worker decisions on task sub-delegation, acceptance, and effort pricing in a distributed manner.
• Simulations confirmed that RTS-P outperforms state-of-the-art approaches in herding mitigation.

Conclusions:

• RTS-P offers an effective individual-level decision support system for HCNs.
• The emergent collective patterns resulting from RTS-P are both productive and robust.
• This approach addresses a critical gap in managing crowdsourcing networks with hierarchical structures.