Finding a Fix for the Most Annoying Technology on the Web
Everybody hates CAPTCHAs.
You know CAPTCHAs – those annoying little boxes you find everywhere on the web with the twisty, distorted letters you’re supposed to decipher and type in to prove you’re a human. They seem to be everywhere - from comment forms to online ticket box offices. They’ve been described as the Vogon poetry of the tech world, and their capacity to annoy and frustrate users is only surpassed by their ubiquity. A 2009 poll by the Daily Telegraph listed CAPTCHAs as the 4th most annoying thing on the Internet, in a close race with Godwin’s Law, pop-up ads, and Rick-Rolling.
CAPTCHAs are evil, but they’re a necessary evil. Their role, quite simply, is to separate humans from bots, and website operators use them for a variety of tasks, including frustrating spammers and ticket scalpers. Regardless of that fact, there’s something fundamentally wrong with the fact that so many applications are dependent on a technology that irritates the hell out of the consumers who are forced to interact with it.
There have been some half-hearted experiments with image technology - like Microsoft's Assira, and some silly ones, like the thing Google was playing with by getting people to recognize if circular images were upside-down, but the only real contribution so far has been Confident's image CAPTCHA – which requires users to click on a sequence of images.
However, that may have changed with a new entrant in the CAPTCHA market. ShareThink Ltd., a tech startup from Ottawa, Canada has just released the beta of a new image-based Turing Test called “VouchSafe”.
VouchSafe has a somewhat different take on Turing tests: instead of identifying text or images, users are asked to make associations between objects, or to find something that doesn’t belong, in a simple challenge reminiscent of the old Sesame Street game, where “one of these things is not like the others”. Users simply draw a line to match two objects or to circle one. The technology works well enough on a computer, but really shines on touchscreen phones and tablets.
ShareThink CEO Chris Ivey says that while his company’s primary focus has been on improving usability, they found that first they had to rethink how Turing tests work.
Up to this point, what we call Turing tests have essentially been mechanical tests of perception – requiring users to identify text or images. The problem is that computers have been getting better and better at doing just that. Even reCaptcha creator Luis Von Ahn has expressed doubts that text-based CAPTCHAs will work for more than a few years*.
“We figured that it’s a bad idea to bet against technology”, explained Mr. Ivey. “Instead, we focused on how humans think about the things they perceive. That’s how we came up with the idea of building an AI that would generate challenges based on the semantic associations between objects. Our goal was to come up with a challenge that was as easy as possible to solve, while still being secure. And we wanted our technology to work without relying on distortion or obfuscation – that’s simply stacking the deck against people who are dyslexic or who have perceptual handicaps.”
Mr. Ivey is a passionate advocate for accessibility and user experience, and wishes there was a more general movement in the industry to replace the much-maligned CAPTCHAs.
“Naturally,” he says, “I want my business to succeed. But more than that, I want to see other people really working on this problem. ‘Good enough’ just isn’t good enough. We have thousands of applications all over the web that depend on a technology that’s only partly effective at best, and that people clearly don’t like. Why is that?”
Why indeed? There may be some compelling business reasons for the current state of affairs, but if so, nobody has done a good job of articulating them. Hopefully other businesses will follow ShareThink’s example, and make a real effort to make the web a little less annoying.