Decision And Analysis Of Crowdsourcing With Various Type Tasks

As a emerging business model in the age of Internet,crowdsourcing has gained wide attention,several well-known companies used crowdsourcing and achieved great success in the business field.The act of companies or institutions outsourcing tasks to an undefined group is selective crowdsourcing,however,because selective crowdsourcing usually uses winner-take-all mechanism in single reward situation,the efforts of most participants except the final winner will just in vain.To explore why this "costly" mechanism become popular in recent years and which type of task is best equipped to this mechanism,in this paper we propose a game model between a sponsor and N participants.The sponsor is to maximize its profit by determining an optimal reward and participants are to maximize their expected payoff by optimizing their effort-spending strategy.In addition,to study whether the sponsor can abandon winner-take-all mechanism in selective crowdsourcing,we assume the total reward is fixed and divided into two classes.Optimizing their effort-spending strategy is still the goal of participants,but the sponsor is to optimize its reward-allocation strategy.In our model,each participant's ability is private,thus all roles in the crowdsourcing system face incomplete information.Our main results are:(i)whether the sponsor can obtain a positive expected benefit is determined by the type of tasks(e.g.,simple,complex or creative task)and the sponsor's efficiency in using the solutions.It should be mentioned that participants' cost functions of different tasks have different patterns.(ii)comparing with simple or creative task,complex task with strong learning effect is more suitable to crowdsourcing platform.(iii)the sponsor's reward-allocation strategy is influenced by the type of tasks,simple and complex tasks are fit for winner-take-all mechanism,but setting two-class reward is better when the task is creative in a special case.Furthermore,we propose several further directions of this research based on our limitations.