Why does the system even care about color in the first place? Can't it measure other facial characteristics, measurements?
Edit: Take a look at this video, it illustrates the problem quite well with a soap dispenser as an example: https://twitter.com/nke_ise/status/897756900753891328
In particular, when using empirical testing to decide if the system is sufficiently reliable, accidentally using an unhelpful sample of faces.
Or even earlier, when selecting the camera and lighting technologies to use.
In color film photography they chose the CYM spectral sensitivities to best represent US white women in heavy make-up under stage lights. It was the 1920/30s afterall. The spectral sensitivities haven't changed much since then as a result (an example of the Lock-In effect in full swing). CCD and CMOS camera dopants are chosen to represent the spectral sensitivities that are similar to film, though physics does come charging in on those.
To be honest, yes, it's hard to photograph darker skinned people without sufficient full-spectrum light. Darker skin absorbs more radiation at the solar-peak frequency (~550nm, the light green of most plants), as the melanin is evolved to do. That said, if your film/sensor can see far outside of the solar peak, then you'll pick up darker skinned people better. Old-timey tin-type photographs could see well into the UV (notice how there never are any clouds in pictures taken way back when) and darker skinned people showed up just fine in those old photographs as a result.
Fun-Fact: the front facing camera on most mobile-phones does not have and IR filter on it, specifically to help with these issues of darker skin. Next time you are near a handle-less toilet with an IR sensor, you can use the front facing camera to see the little IR LEDs beeping to see if you are there. You can also use the front-facing camera to see security lights and what not, typically those are IR ones too.
UV light could also be used, but it is more damaging over time, and thus lower frequency IR light is preferred.
Also could easily be a result of biased or limited training data.
Why is it an acceptable basis to discriminate on height, but not melanin content?
Because there’s a ton of “one size fits all” products which don’t fit the ends of the bell curve, and reflect the choice of manufacturer to not serve all body types.
I think it’s a poor choice for a door lock, but I’m not sure I a priori think it’s unacceptable for a product to not work for some (~10%) portion of the population — especially when it reflects an engineering reality (skin reflectivity; feature sharpness), and not choice.
Because society as a whole hasn't spent hundreds of years literally enslaving short people before deciding that strict moral and legal standards are necessary to prevent it from happening again.
Actually, they have done exactly that. For many centuries before appropriate nutrition became essentially ubiquitous, there was indeed a highly-significant correlation between height and social class - and the lowest class often faced near-slavery conditions in surprisingly recent times, especially outside Western Europe. Of course, you're right that "strict moral and legal standards" are rather lacking on the issue. I'm not sure that we would agree on whether that's a desirable state of affairs.
If my employer specifies a uniform, it's their obligation to provide one to 100% of their employees, not the 90% covered by the cheap "one size fits all" company they prefer to use.
If my landlord specifies a locking system, it's their obligation to provide one that unlocks 100% of the time for everyone who rents there.
There's also an issue of scale: an apartment is a fairly major decision with a lot of other factors, that affects a huge chunk of your life. Being locked out by a small, easily changed decision on the landlord's part is unreasonable. Conversely, if you can't wear a "one size fits all" product, there's plenty of other options out there, and you're not potentially creating segregated neighborhoods, locking black people out of living near good jobs, etc..
Only if such clothes are reasonably commercially available — otherwise, they have a bonda fine business reason to discriminate in hiring, and may do so.
I think you’d find you also had a really hard time for suing if they just didn’t hire you on the basis of the uniforms they had not fitting.
So you seem factually incorrect about the height case.
> if you can't wear a "one size fits all" product, there's plenty of other options out there, and you're not potentially creating segregated neighborhoods, locking black people out of living near good jobs, etc..
This is again mistaken.
If you’re at the ends of the bell curve for height, you can’t reliably find work clothes through normal channels, and are thus unable to reliably signal your qualification for a white collar job.
The lack of height appropriate clothing impacts the ability of very tall or very short people to “dress for success”, and hence harms their chances of attaining the best jobs.
Your comments merely reflect your cis-height privilege and implicit heightism.
Further, it speaks to your implicit racism, because you view melanin as somehow biologically more meaningful than height.
Or at least less visible light. I wonder how skin pigmentation affects imaging with infrared light?