Joe Gregorio’s clever alternative to CAPTCHAs

You’ve probably read some articles talking about how overzealous websites make CAPTCHAs with distorted text that are really hard to read; e.g., #12 on Guy Kawasaki’s list of ways websites hinder their own market adoption. Some people have thought of alternatives to this: in Dean Allen’s Textpattern CMS, you have to review what you’ve typed at an intermediate screen which will defeat the basic automatic form parsing that spam bots are often programmed to do.

Without making you do mental arithmetic, he just asks you to put your name in a certain format:

You must enter your name followed by something in parentheses, either an email address or the URI of your web site. For example: joe (http://bitworking.org)

If you don’t enter it correctly then your comment won’t be accepted. (Hey, it’s better than making you squint at a stupid little captcha image trying to figure out the numbers and letters.)

It requires you to stop and think for a moment, but you really should stop and think for a moment if you’re going to comment on a blog entry anyway.

The essential idea of a CAPTCHA is to ask a user to carry out a hard AI task, and it need not only be visual pattern recognition; I think natural language parsing is a good candidate because we really can’t do a good job of things like machine translation and so on. The hard AI problem here is having you understand how to format data from plain text, and yet a textual task will still be compatible with people using screen readers. It would also be easier to implement for the web developer; many frameworks have basic support for input validation. Gregorio’s format would only require a straightforward regex, rather than a call to a special image generation library. And he is correct in pointing out that visual CAPTCHAs discriminate against people who have poor eyesight.