Google reCAPTCHA: "Tough on bots, easy on humans" ???
The first lecture made me think of a story I've overheard from one of my friends. He told me a story about his experience with Google reCAPTCHA. reCAPTCHA is an advanced risk analysis engine is a free service offered by Google to any website. Website designers can choose to use this service in a way to protect their selves from non-humans (AI) entering or filling in their online questionnaires (for example).
Anyhow, he was trying to online reserve a time slot for certain event. However, this event is highly popular. So popular that he was even forced to set his alarm in order to be on time to select a time slot. Once you succeed in selecting your time slot, his reservation was not comfirmed yet. He still needed to enter his name and perform the reCAPTCHA task displayed on the left. Somehow, he always lost the reservation game on performing this specific task. Because of the stressful situation he always took way to long to proof that he is a human, leading into multiple try-outs before a successful reservation.
This made me think about how the developers of Google totally miss the objective of reCAPTCHA. The engine aims to be "though on robots, easy on human", however if a human is at its most human (by being stressed out) reCAPTCHA becomes "though on robots, though on human".
Robots stressing their asses off.
Imagine already being stressed and in a hurry and then having to do this vague reCAPTCHA task. You might ask yourself do I only need to select the traffic lights? Or also the construction were the traffic lights are attached to? Do I only need to select the first two or does that one in the back also count?
These thoughts are human but can only be confirmed by guessing. However guessing in stressful situations often leads to wrong guesses. Leading to an annoying situation were the user knows he is human, but not being able to prove it. I'm interested to investigate how you could implement/visualize this process of stress and frustration in a robot.