Hacker News Reader

Comment by tj-teej | Hacker News Reader

A lot of people will say "this is a limitation of Computer Vision", but they are missing something.

If these kinds of technologies treat the hair of a certain race as a P1, then the technology (and people who designed the tech and the program) are participating in discrimination.

I'm not saying this to be 'woke' or to 'cancel' TSA. It's just a fact; and once we can all admit it (a lot of people in this thread don't seem to be able to), then we can make changes to fix it.

Editorial: If an aspect of traveling makes you feel like an 'other' in your own country, that's wrong, and I don't think public funds should be encouraging that. Public buildings have to have wheel-chair ramps, and TSA body scanners shouldn't be racialized.

View on HN

Why does it need fixing per se? (Assume optional plurals)

If the parameter is incorrect, then the parameter should be adjusted correctly. Note that having an availability bias in the dataset is also incorrect, the parameter will reflect the incorrectness.

However, if the parameter is adjusted correctly: what is the problem other than people not wanting to be confronted with uncomfortable truths?

Read the article again. The machine has a problem with certain hairstyles.. not hair of certain races.

People of African descent have a propensity towards the problem hairstyle that skews the results.

If your argument is for 'de-racialization' then you need to convince people to change their hairstyles to make everyone more similar.

You can't blame a machine for treating a 1 as 1 and 0 as 0.

You're right you can't blame the machine. I'm blaming the people who designed it, who defined what the 1 and the 0 are.

If everyone in the country had the "certain hairstyles we'd have no problem saying "this machine doesn't work".

And frankly I don't think anyone who knows a lot of African-American Women would even consider asking them to change their hair.

> If everyone in the country had the "certain hairstyles we'd have no problem saying "this machine doesn't work".

On the contrary, it would be easier to say that the machine does work because it does flag down people who have suspicious body features regardless of race.

You can't blame the people that designed it.

Some hairstyles are just bunched up and the machine can't tell what it is, so it gets flagged for manual inspection.

This is not a problem that is easily solved with technology.

Your only options are:

(1) not checking anyone's hair

(2) manually checking everyone's hair

(3) letting the machine do it's job and accept the fact that it can't scan certain hairstyles.

My understanding is that they are not targeting a certain type of hair, rather a certain type of hair has characteristics that the machine cannot distinguish from potential risks.

If all hair had that problem then we'd say the machine doesn't work.

They launched this new tech knowing a "certain type of hair" (specifically black women's) triggers the machine (and triggers degrading searches). I'm saying this machine shouldn't be launched if that's the consequences, I'm saying that it's discriminatory.

Reply on news.ycombinator.com