Crowdsourcing has become a hot topic in recent days due to its potential in processing human instinct problems. Human intelligence, as a type of computing processor, outperforms the machinery computing processor greatly in terms of processing abstract and incomplete information. However, the uncertainty of human workforce has strongly constrained the application of crowdsourcing in larger aspects. Generally speaking, the uncertainty of human workforce results from two main factors: one is human’s diverse capability towards different kinds of tasks; the other is the existence of malicious human workers. For these reasons, the result of crowdsourcing should not be accepted directly, and it is critical to create a set of mechanisms to guarantee its credibility.

To cope with the a...[ Read more ]

Crowdsourcing has become a hot topic in recent days due to its potential in processing human instinct problems. Human intelligence, as a type of computing processor, outperforms the machinery computing processor greatly in terms of processing abstract and incomplete information. However, the uncertainty of human workforce has strongly constrained the application of crowdsourcing in larger aspects. Generally speaking, the uncertainty of human workforce results from two main factors: one is human’s diverse capability towards different kinds of tasks; the other is the existence of malicious human workers. For these reasons, the result of crowdsourcing should not be accepted directly, and it is critical to create a set of mechanisms to guarantee its credibility.

To cope with the above challenge, we come up with the model of trustworthy crowdsourcing, which guarantees the crowdsourcing result satisfying specific credibility requirement. Our model consists of two fundamental modules: the credibility tuning module and the variation detecting module. The credibility tuning module adds necessary redundancy to crowdsourcing based on the estimation of crowd’s capability, so that the credibility requirement for the crowdsourcing result is satisfied. The variation detecting module monitors the change of crowd’s capability and adapts the execution of crowdsourcing, so that the credibility is maintained.

Extensive experiment study is made to verify the performance of trustworthy crowdsourcing. According to the experiment result, the proposed methods are effective in real world application.