hp's face tracking software problem or racist computers?
I'm sure most of you have heard about the video “HP Computers are Racist” by now, Desi and Wanda have spoken to Mashable about how they discovered the flaw at work and put the video out in a humorous manner to get other people's reaction and insight on the flaw they discovered.
Desi has been accused of trying to get a free ride by some commenters but they missed the part when Desi says he already bought one for his wife, and it was probably for Christmas. He now has a product with a feature he can't utilize.
The problem with the software is a complexion or skin color issue, which can be attributed to race and clearly a sensitive topic. The reality is we know that an inanimate object has no way to determine race but the programmers decided to use algorithms to detect features that may be common in some ethnic facial features but not others.
Laptop Magazine decided to test the HP Envy 15 with three different facial hues. I noticed the camera react differently depending on the facial features of who was sitting in front of the webcam. HP's statement in reaction to Desi's and Wanda's video states:
“The technology we use is built on standard algorithms that measure the difference in intensity of contrast between the eyes and the upper cheek and nose. “
As many of us know some ethnic groups have darker hues, higher cheek bones and/or broader noses that can affect the “intensity of the contrast” to produce various outcomes. You can see this in the first video by laptop magazine as Dana, a white female sits down, the camera reacts immediately, when K.T., a brown female sits down, the camera reacts, although a bit slower. When Jeffrey, a black male sits down, there is no reaction at all. [note: that K.T. is also wearing glasses that are casting a shadow in the region between the nose and cheek.]
“On any given day, I might collaborate with HP employees in regions ranging from Japan, India and Latin America”
I am sure their teams are diverse but there are some ethnic features that are commonly found in descendants of Africa, Haiti and a few others that were obviously overlooked in testing the product. This is one of the reasons the flaw wasn't caught before the product was put on the market. As Laptop Magazine states:
“when testing this software, how many of the involved project members were dark-skinned? How many different lighting conditions were tested? ….... Perhaps this incident is an indication that the software makers should involve a greater pool of testers when designing these features."
As we continue to move into more social technology we will find a variety of product errors when certain cultures are missing from the design teams or during product beta testing.
HP's recent experience should be a helpful hint to other companies. Look around at your teams and check if you have enough diversity to test your product. If not just reach out to other various networks, you will be surprised at the response, and as a result improve the quality of your product.
This incident has caused an interesting stir in social networks. If you want to follow some of the conversation on twitter use or search the #HPFail hash tag.
For myself, it has been extremely helpful for my projects to have documented samples that help explain these types of issues in sensitive discussions about team diversity, particularly in tech.