Traffic Flow explores a new technology known as "crowd sourcing" in which large volumes of small tasks are distributed to human workers. Some common examples include Wikipedia, eCommons, and Amazon's Mechanical Turk. These peer production systems are akin to the distributed computing efforts of groups such as SETI@home or Folding@home, but instead of using computers' free cycles they use humans' free cycles. These systems manage to produce massive output (c.f. Wikipedia) by dividing labor into such tiny fractions that people often contribute for no pay at all.
In this experiment, we are exploring the Mechanical Turk, in which people participate for "micropayments," typically on the order of a few cents. It is an open question why people are participating and what kinds of tasks can be accomplished in this framework. For this task, participants sort traffic images into chronological order, something that would require large amounts of image processing for a computer but is very easy for a human. This allows means a large number of images can be sorted much faster and more accurately than would be possible otherwise.
In the future, we intend to try other similar tasks, such as characterizing traffic, identifying vehicles in images, or determining if the same vehicle appears in multiple traffic images.
Traffic Flow was created by the Economics and Social Systems group.
Even though you are using the mechanical turk for crowd sourcing, what about the (micro)payment model for the crowd-sourced users ? Do you think a viable business model can be built out of requesting people to identify traffic characteristics ?Comment by gangadhar npk (November 24, 2008, 7:50 pm)