Like other controversial proposals, a plan for protecting polar bears threatened by receding ice touched off a torrent of mail--more than half a million comments--to the U.S. Fish and Wildlife Service (FWS).
This time, to head off a logistical nightmare, social scientists and computer researchers funded by the National Science Foundation are creating language-recognizing algorithms to spot form letters, group similar comments, and even determine whether a comment is pro or con. The hope is to enable bureaucrats to sample relevant letters without having to plow through all of them.
But teaching computers to get the gist of a letter isn't easy, says political scientist Stuart Shulman of the University of Pittsburgh in Pennsylvania. "People will say 'I hate the Bush Administration,' … but they are for the listing." He adds that hundreds of thousands of emotionally charged form letters from environmental groups create "noise" that can drown out what the agency wants to hear about most: science and economics. FWS has until January to make a final decision.
