Google Acquires reCAPTCHA: Potential Boost for Web Accessibility

Google

Web Accessibility activists take note: Google has just purchased reCAPTCHA, the most widely used audio CAPTCHA on the internet. In this post you can read about

Simplified Summary of this Document


Google reCAPTCHA Announcement

The September 16 Google blog post announcing the reCAPTCHA purchase did not specifically mention the audio aspect of the ReCaptcha product, or the critical importance of CAPTCHA alternatives to blind and visually impaired computer users. Still, the short post, signed jointly by the Google Product Manager and ReCaptcha’s co-founder, did say this:

Improving the availability and accessibility of all the information on the Internet is really important to us, so we’re looking forward to advancing this technology with the reCAPTCHA team.
— from Google Blog Post, September 16, 2009

I’m thrilled that Google wants to improve the accessibility of all the information on the Internet. A big step toward that goal is making sure that ReCaptcha’s audio functionality is robust, fully usable and accessible. I hope the ideas here will help Google in that effort.

Blind User Testing for reCAPTCHA

Google would not be the successful company it is if it had not mastered the art of user testing and consumer input. Those skills must be put to use with the newest member of Google’s stable.

In the next sixty days, Google should convene a diverse group of blind and visually impaired computer users specifically to test and give feedback on ReCaptcha. Google should continue working with that group and other blind and visually impaired computer users as it moves quickly to implement suggestions and strengthen and improve the reCAPTCHA product.

Because neither traditional visual or audio CAPTCHAs are effective for people who have both impaired hearing and vision (an issue for many elders and others) additional user testing for this group would help establish a more broad based alternate CAPTCHA. (Logic CAPTCHAs, like the one on the Contact Page of this site (asking the user “how do you spell law”) are present on a few smaller sites, but don’t currently have the security protections needed by most sites worried about significant bot trouble.)

Key features of effective audio CAPTCHAS

Here are some key aspects of a successful audio CAPTCHA, some of which reCAPTCHA has already done well, some of which still need work. Thanks to Jim Thatcher, Brian Charlson, and other ACB members who have worked with Linda Dardarian and the Law Office of Lainey Feingold to improve audio CAPTCHAs on various websites during Structured Negotiations efforts. (A complete list of web accessibility agreements reached using the Structured Negotiations process is available in the Categories Section of this website.)

  • Clarity of Spoken Words. If the printed letters of a visual CAPTCHA are not squiggly enough, the CAPTCHA won’t work to keep out the malicious programs that are the very reason for the CAPTCHA’s existence in the first place. On the other hand, if the letters are too indecipherable, legitimate sighted users will not get past the gate.The same is true for the audio version. Just as sighted web users must be able to make out the visual words, so too must blind web users be able to decipher the spoken words. If those words have been too garbled in the effort to keep out bots, blind and visually impaired users will be locked out of the very content that the CAPTCHA is designed to protect.The current audio version of reCAPTCHA, with radio clips instead of random, masked, numbers, is much improved over previous versions. Still, the failure rate for people listening to the audio CAPTCHA must be no higher than the failure rate for those viewing the visual CAPTCHA. Anecdotal information shows that this benchmark has not yet been achieved.
  • Volume of Spoken Words This may seem obvious, but an audio CAPTCHA cannot work unless the computer user can hear it. Appropriate default volume settings as determined by rigorous user testing are critical.
  • Clear instructions. Instructions for using the audio CAPTCHA must be clear and consistent. For example, if the instructions say “please type the words” and then digits are read, it is unclear to the site user whether s/he is supposed to enter four digits, or spell out four numbers. Different instructions must be given for digits than for word strings. If Google continues reCAPTCHA’s use of both word strings and digits, clear instructions would be something like “If you hear numbers type the digits without spaces. If you hear words, type the words with spaces between them. Don’t worry if you have trouble with one word. Just enter your best guess”
  • Tab Order Visually impaired computer users, and computer users with other disabilities as well, do not use a mouse to navigate a web page. They use a keyboard which makes correct tab order a critical issue for the accessibility of the reCAPTCHA product.
  • Proper Label Elements and Title Attributes. To prevent screen reader software from searching for text that may give a confusing (or flat out wrong) message, the title attributes and label elements must be carefully and precisely used. This is especially important on the form control. When the audio is being used, the title attribute should be something like “enter the words you hear”.
  • Ability to [easily!] repeat the audio. Just as the squiggly letters of a visual CAPTCHA are often hard to make out for sighted computer users, the audio clips that are at the core of reCAPTCHA’s audio program sometimes take a couple of tries to decipher. Wisely, reCAPTCHA has a “play the sound again” link, but in many current implementations, that link does not take the user back to the entry field to type in the words. The script must automatically put focus on the input field after the user opts to hear the sound again, rather than requiring the user to find the input field again.
  • Help and Information Any technical support provided to Google’s CAPTCHA users, including simply on-line instructions, must clearly explain how the audio CAPTCHA works and what the user should expect. Instructions for sighted and blind users must be clearly described: use of the audio CAPTCHA cannot be listed as an afterthought at the end of a list otherwise geared to the sighted user.

Congratulations Luis von Ahn

Although reCAPTCHA has had some problems with its audio version, I want to congratulate Luis von Ahn, the cofounder of reCAPTCHA, and recognize the important work that has already been done to improve the audio product. As a result of various Structured Negotiations efforts involving the accessibility of websites, I contacted von Ahn in March of this year about the importance of audio CAPTCHAs for blind computer users. He has been willing to listen, open to our suggestions, and has made many improvements. It is good news that von Ahn went to Google along with the product he developed while at Carnegie Mellon. With the weight of Google now behind the most prevalent audio CAPTCHA in the world, a fully accessible and usable audio CAPTCHA will hopefully be here soon.

Simplified Summary

This is a post about Google. Google just bought a company that makes it harder for computers to infect a website. The Company is called ReCaptcha. Blind people can listen to words and then type what they hear. Those words are called an audio CAPTCHA. This allows the computer to know that a person is using the website, and not another computer. This post has ideas to make it easier for blind people to use Google’s new product. [back to Google Announcement ]