Missing aviator and adventurer Steve Fossett is lost somewhere in 10,000 square miles of Nevada desert. A traditional and comprehensive rescue effort has been in progress for days, but it is difficult to know where to look for him because he did not file a flight plan.
An unorthodox group of searchers are approaching this problem from a non-traditional angle, using updated satellite imagery from Google Earth and a new Web service from Amazon called the Mechanical Turk. They hope to harness the wisdom of crowds to examine in detail images of all 10,000 square miles in hopes of locating the crash site and effecting a rescue of Mr. Fossett.
The Mechanical Turk (MTurk) uses computer intelligence to collect, analyze, and present the wisdom of the crowd, which Amazon terms human intelligence tasks (HITs), into an “artificial artificial intelligence.”
MTurk is named after a turbined wooden mechanical chess-playing automaton from the 1700s that could beat most human players. Its creator would open doors on the cabinet that the mannequin sat on, revealing an elaborate set of gears. But it was a hoax. A Turkish chess master was hidden inside the cabinet and he made the moves for his mechanical counterpart.
What does all this have to do with reporting? I suspect MTurk is one possible solution to problems that open source and crowdsourced reporting experiments like offthebus.org or the Sunlight Foundation encounter with managing, analyzing, and preparing data as news.
Today, computer processing power is often considered to be faster or more efficient than human thinking. In MTurk, Amazon is working from the idea that HITs can be crowdsourced and that application programming interfaces (APIs) can effectively organize how human intelligence is collected, organized, and presented.
The reliance on the “wisdom of crowds” is a similar proposition — but efforts to harness the power of crowdsourcing can bog down when too much data comes in with little or no way to manage or interpret it. A Web service like MTurk might help solve this knowledge-management conundrum.
When MTurk isn’t working on life or death problems, Amazon intends to make money with it. They aim to match the needs of a company or business with “thousands of high-quality, low-cost, global, on-demand workers — and then programmatically integrate the results…” of the work into a customized output. This may not be happening yet, but using MTurk’s to organize volunteers in a sort of “gift economy” situation is interesting. Some suggested uses for MTurk include:
- People can ask questions via computer or mobile device, and workers return answers.
- Writing reviews, descriptions and blog entries.
- Finding specific information in large legal and government documents.
Think about the last time you called customer service. Where in the world was the customer support person you talked to? No doubt this kind of service will transform some forms of work.
For crowdsourced investigations, the workforce often consists of volunteers — similar to the search for Steve Fossett. With MTurk providing the data backbone or information scaffold for an investigation, the “pro” part of a “pro-amateur” journalism team might be paid to write stories about what the crowdsourced project reveals.
Returning for a moment to the unfortunate Mr. Fossett’s predicament, Google Earth is providing images of the earth from space to the MTurk Web service, which in turn is packaging thousands of images into a form that allows willing humans to bring their intelligence to the search.
Follow or join the search for Steve Fossett via MTurk.
