Participation in crowd systems

There is growing interest in using crowds of human workers to perform information processing tasks that are difficult for machines. At IBM, we have deployed a large internal crowdsourcing contest system for software development, which has now successfully completed thousands of tasks. Contrary to concerns of power efficiency and bandwidth in machine systems, player participation and motivation are key concerns in such crowd-based human computation systems. In this paper, we demonstrate that player participation is well-described by a power-law distribution, much like Lotka's Law of scientific productivity. We further describe a momentum-based generative model for participation that explains the observed empirical phenomenon and provide a thermodynamic interpretation, discussing what it means for a crowd system to heat up. Overall, player motivation is essential for driving participation and ensuring a reliable delivery platform; several intrinsic, extrinsic, and social motivators have been identified.