Skip to main content

Information Cascades and Crowdsourcing

Crowdsourcing is a phenomenon that uses the “wisdom of the crowd” to perform tasks that would be onerous for an individual to do alone. However, the openness of problems solved via crowdsourcing makes them especially vulnerable to effects of information cascades.

Information cascades can have a positive effect on crowdsourcing. For instance, in the 2011 Darpa “Shredder Challenge,” groups competed to put together five shredded documents (in varying degrees of shredded-difficulty). Most groups used powerful image-processing software, but Iyad Rahwan and Manuel Cebrian of MIT chose to enlist in crowdsourcing. With the help of thousands of people, they were able to successfully piece together parts of several documents. The effect of thousands of collaborators spurred individuals to continue working on the project. However, a negative effect of information cascades was observed on day 5, when a saboteur infiltrated the project and managed to undo 5 days of work. While this didn’t spur collaborators on the project to undo work they’d previously done, it did diminish morale and cause the team to lose the lead they’d built up.

Another example of the cascade effect in crowdsourcing is apparent when groups of people work together to identify unknown persons. The “Tag Challenge” (2012) was yet another Darpa challenge in which groups had to seek out five actors disguised as jewel thieves hiding in five European and North American cities. Rahwan and Cebrian found crowdsourcing extremely effective here – they created an app that allowed participants to quickly upload pictures and locations of the “jewel thieves” so that others in the area could aid in their capture; the cascade effect can be seen here in a small group of users spreading information to a larger group that causes the larger group to act in a certain manner. However, the opposite – incorrectly identifying individuals – is certainly possible. This occurred in the aftermath of the 2013 Boston Marathon bombings, when users on Reddit incorrectly identified the bombers, which led to innocent individuals being harassed. The belief of a few led many to believe (incorrectly) that certain individuals were the bombers, which certainly was not the case. Thus, the effect of information cascades on crowdsourcing is paradoxical in that it can simultaneously be used to benefit or impair the judgement of the crowd.



Leave a Reply

Blogging Calendar

November 2014