reCAPTCHA is UX’s worst nightmare
or: How to recognize the unrecognizable
I get it. The machines are rising, and soon, we’ll all have to face being fired because a new designer-bot designs better interactions than us.
It started when the factory workers were replaced by machines, through Tinder killing the matchmaking business, and now this:
This GIF is not that new, but it gives me a great segue to my subject.
We’ve been fighting robots for a really long time. CAPTCHA started its journey at the beginning of this millennium. You know what that means — CAPTCHA is a millennial, and similar to the other millennials — I kinda hate it.
Let me just start by talking about the acronym. CAPTCHA stands for Completely Automated Public Turing test to tell Computers and Humans Apart. So, it’s actually more of a CAPTTTCHA.
Acronyms aside, CAPTCHA started as a way to prevent automated form filling, by giving people a picture of a distorted word (often a really weird one) that they need to type correctly. The idea behind it is, of course, the fact that a computer program wouldn’t be able to understand the word when it’s distorted like that.
The thing is, sometimes humans fail to understand the words too.
Throughout the years, people in Google understood humans’ ability to recognize words has a great value to the ever-expanding AI technology, and reCAPTCHA was born. reCAPTCHA had two words. One of the words was the same as a distorted word in CAPTCHA, and the other was a scanned word from a book. The assumption was that if people understood the first word, they also understood the second word, and so the computer was actually learning from the user’s answer. Again — the machines are just becoming stronger, while we’re busy typing weird words.
And then came the image recognition
God have mercy, the image recognition.
The image recognition reCAPTCHA gives you a grid of images, and ask you to select all the images that match the description. Sounds legit at first. Google advertises this tool as an easy tool for humans, that shouldn’t be hard at all, but in the reality, this test is just hard and frustrating. A couple of days ago, I spent about 5 minutes on this, trying to pass the test.
On the left you can see the way Google is advertising reCAPTCHA, and on the right is the more common reCAPTCHA.
And of course, this image recognition is helping Google’s AI to recognize objects.
The thing is this: I see why it’s a great opportunity to use people to teach the algorithm, and I also see why we must have a strong protection against any kind of an automated attack, but why, oh why, do I need to spend so much time recognizing a word that is so distorted, even its mom wouldn’t recognize it (I’m automatically assuming every word has a mom)? And why should I use all my mental resources to recognize a storefront, or a hidden stop sign in a series of pictures?
I feel like this is getting a little out of control. We’re teaching computers to think like humans, by making humans burn their brains trying to recognize abstract images, and I think this is making people frustrated in a horrible way.
Alternatives?
If we’re helping Google’s AI, I don’t see any way we can outsmart an algorithm, but we can at least make it look like we are. We can, for instance, use Google’s “Quick, Draw!” interface, and let people draw things:
Or maybe even talk about emotions — a thing that most people can recognize quickly?
In conclusion
I’m not sure what’s the ideal solution for a good reCAPTCHA UX, but I have to say this reCAPTCHA thing really got me missing the old CAPTCHA, or as I call it — CAPTTTCHA.
