cell phone

Why Does Facial Recognition Software Not Always Recognize Dark Skin Color?

If you are a person of color, you most likely have experienced issues when it comes to facial recognition software.  I mean, look at how many people were running into issues with FaceID.  Don’t get me wrong, it’s not only limited to facial recognition software.  Any technology that relies on detecting a skin color can be problematic if the technology isn’t designed properly.  We saw this a year ago (or more) with automatic hand dryers in washrooms not detecting dark colored skin.  We’ve seen this with apps as well – especially the ones that are trying to distort your face into a funny looking image.  Microsoft is now saying that they’ve improved their facial recognition software so it’s no longer racist.

Earlier this year, they released new software to developers, and while some people took this a bit too far, the software itself didn’t work the way it was intended.  Microsoft’s Face API offered developers access to what they claimed to be advanced facial recognition technology that could be integrated into their own products.  But researchers at MIT and Microsoft’s own New York-based research lab found that the company’s new technology not only had some big issues with race but also gender.

Facial recognition software

According to reports, Microsoft’s Face API facial recognition technology was remarkably proficient at identifying white males. In fact, the team found that Microsoft’s tech could identify white male faces with an error rate of 0%. I want to say that this is great, but I think I need to highlight the fact that it was identifying white males.  When it came to black women, however, that error rate went to 20.8%.  Which means, the software wasn’t going to work the way that Microsoft had intended it to.

But on Tuesday, Microsoft announced that they’ve updated their facial recognition technology and made some significant improvements in the system’s ability to recognize gender across skin tones.  John Roach writes the following on their AI blog:

“That improvement addresses recent concerns that commercially available facial recognition technologies more accurately recognized gender of people with lighter skin tones than darker skin tones, and that they performed best on males with lighter skin and worst on females with darker skin. With the new improvements, Microsoft said it was able to reduce the error rates for men and women with darker skin by up to 20 times. For all women, the company said the error rates were reduced by nine times. Overall, the company said that, with these improvements, they were able to significantly reduce accuracy differences across the demographics.”

iphone x

I think this improvement is great, I do wonder what kind of testing that is done in advance with a variety of skin tones.  What’s interesting to me is that this keeps happening.  And the fact that there are so many instances of technology not working with specific skin tones is very telling, don’t you think?  It could be one of two things – the technology is inferior, which is sometimes the case, or the testing protocol isn’t diverse enough.

When it comes to the technology itself, we have seen instances where inferior technology was used, and that’s why it couldn’t identify skin that wasn’t white.  In that particular case, it was related to lighting, but the fact remains that this keeps happening over and over again.  While I am not trying to be particularly hard on Microsoft, I do think that everyone needs to be more cognizant of how their technology will impact people of color.  Whether that’s exclusionary by nature of the fact that they have a darker skin color.  Or, as we’ve seen in media and entertainment, a lack of diverse representation can have negative impacts on an individual’s well being.  Even though these things aren’t intentional, they are still happening and that needs to stop.