Interesting toy, using the ML tools against themselves. It deliberately shows you how broken these surveillance algorithms are (e.g. not recognising PoC), how superficial and biased the judgments are (assuming slim young white cisgender people who trust sites with personal info are the norm), and how horribly they can impact the “non-normal” (not “Normal” = potential shoplifter or terrorist). I’m glad that the EU is taking these matters seriously and exposing its citizens and the wider world to their dangers in a creative way that respects privacy.
FWIW, I got a score of 75% Normal (“Violently Average”). It would have been higher if I didn’t self-report as an Old (“Normal” range of respondents is age 12-32); if I wasn’t assumed to be lying about my age (the algorithm flatteringly assumed I was younger than I am and also made the unusual assumption that I was pretending to be older); and because I wasn’t wildly darting my eyes around the screen and moving my mouse and switching tabs (apparently “Normal” people are more prone to do this).
Of course, no-one who actually knows me would consider me fundamentally “normal” (I usually get “eccentric”). Also, a few of the results that put me in the “Normal” range were likely errors. For the techbros and marketing gurus and security-state apparatchiks who set the Western parameters of “Normal”, though, I seem to pass well enough to go unnoticed.
