Crowdsourcing Workflow

Crowdsourcing workflows, which break down complex tasks into smaller units for distributed human completion, are being re-examined in light of large language models (LLMs). Current research focuses on adapting crowdsourcing techniques to design and improve LLM-based systems, leveraging established strategies for handling errors and ensuring diversity in outputs. This involves exploring how LLMs can augment or even replace human workers within existing workflows, assessing their relative strengths and weaknesses across various sub-tasks, and identifying optimal human-LLM collaboration strategies. The ultimate goal is to enhance the efficiency and quality of complex tasks through hybrid human-AI approaches, impacting both the design of AI systems and the future of human computation.