Last Sunday’s New York Times reported on captchas, the ubiquitous website security tests that require users to decipher and enter words or letters presented in squiggly type. In theory deciphering the puzzle requires the application of human intelligence, frustrating automated programs that would access the site for malicious or mischievous purposes. As the ability of automated programs to crack the puzzles becomes more sophisticated the captchas must evolve to outwit them. Some captchas require a user to orient a “randomly rotated” image, for example. Most of us have experienced a captcha with indecipherable letters. Is it an i? A j? A t? An f? Sometimes I reach for stronger reading glasses and position my face inches from the screen, muttering “who created this chicken scratching?” in frustration.
I was inordinately pleased to learn that this frustration serves a greater cause than site access, at least for reCaptchas (which this blog uses for new-user registrations). reCaptcha uses non-machine-readable words from books that are being digitized. If a scanned book page is wrinkled the scanner may return gibberish that requires human interpretation. reCaptcha presents those words to users who parse the mangled letters in the course of logging into a site, then reports the results to those converting the books to digital form. If 6,075 people decipher the word as “constabulary” then “constabulary” it must be. reCaptcha’s clever efficiency will temper my frustration the next time I fail one of its tests for human intelligence.