Project Description


CAPTCHAs are systems that pretend to tell humans apart from bots. They are dispatched all over the internet. We find CAPTCHAs on thousands of websites, webpages and blogs. CAPTCHAs secure websites, forms and databases from malicious bots. But they are not only about segregating bots. They are also part of one of the most important enterprises of labor extraction and human cognition exploitation: each time you take on a CAPTCHA challenge, you work for free for Google Inc. and take part in the training of their algorithms. The first generation of CAPTCHAs, text based, was designed to train text recognition algorithms, while the second generation was designed for the training of computer vision algorithms. In addition, like the Facebook buttons, they are harvesting all the data from the websites where they are hosted. The most recent CAPTCHAs are invisible. They silently analyse the behavior of a user during his visit to a website, to determine if he’s a bot or a human. Depending on the user’s interactions with the website, he is assigned a score, that varies from 0.1 to 0.9. The lower scores (0.1) classify the user as a bot, while the greater (0.9), classify the user as human. A webmaster can adjust the threshold, so to segregate bots at a certain score, usually set by default at 0.5.

As those systems became more and more perfected, one can wonder if we can even fool them. DOUBLE NEGATIVE CAPTCHAs provides a training session, where the metrics of the system are made visible, in order to train ourselves to fool the system.

If we are slowly turned into bots when we use these systems, can we regain our humanity by camouflaging ourselves into bots?

Let’s try

Paris (France)
7 August 2020