21 - 08
You can apparently be too black for some automatic devices to properly serve you; the latest incident of a machine being ‘racist’ shows a soap dispenser seemingly refuse to dispense some soap to a rather patient black man.
The video, shared by Nigerian Chukwuemeka Afigbo, the head of Platform Partnerships for Facebook (MEA), begins with a white man getting soap from the dispenser without any problem; he places his hand under the dispenser, and it does its job.
For the black man however, despite more than ten seconds of staying his hand, waving it around, swinging it sideways, the dispenser is unresponsive.
The machine however responds promptly, twice, when the black man takes a piece of white tissue and holds it beneath the dispenser.
For some social media users, the video is only demonstrating what happens when the ambient lighting is poor, but many others are convinced that racial diversity in the dispenser’s design phase could have probably averted this apparent problem.
After all, this isn’t the only device that has demonstrated such ‘racist’ responses: in 2009, HP had to deal with accusations that its face tracking webcams couldn’t detect black faces.
Facial recognition softwares have also been found to be worryingly inaccurate in detecting black people, and a beauty algorithm chosen to select the most attractive ladies in a pageant chose only one black lady among its 44 eventual winners.
Similarly, a research conducted by an MIT team found out that quite a few language processing tools, when given tweets containing African-American slang, classified such tweets as Danish.
Part of the problem contributing to such incidences is that the tech industry skews heavily towards white and male; consequently, during the development phase of products, products are likely to be tested on white males, and ultimately refined to suit this demographic.
It’s not a uniquely Caucasian thing: face-detecting algorithms developed in Asian countries have been found to be highly accurate in detecting Asian faces, while faring poorly when presented with non-Asian faces.