The authors use of language within the first couple sentences exposes what seems at first glance like a disdain for the people who live in these houses.

This paper provides a comprehensive way for creators of ML systems to classify their systems and could hopefully lead to more transparency and credibility within the AI Community.

"There are certain situations where we should not be predicting with a ML system at all." In tech today a lot of effort is given to minimize cost and maximize effectiveness which is why these shortcuts are often taken with ML. What kind of incentives could convince companies to actively look at individuals instead of screening passively with AI.

Would the introduction of more care for minority data groups decrease the overall accuracy? If it doesn't why haven't companies worked to do this, and if it does how can this be reconciled with more fair ai systems.

Health care and other important industries have algorithms that are not open source or disclosed to the public. I wonder what effect open sourcing some of these would have.

Since labelling data sets is inherently human intensive I wonder what could be done systematically to try and reduce these biases in the future.

Just 3% of photos containing black faces, in a dataset that is being widely used to train neural networks is very problematic and representative of a deeper problem within the data science community

The entrenched racism within our society really has such a strong effect on technology. These examples really illustrate the new jim code.

As someone studying computer engineering I have tended to look at algorithms as these unbiased machines but now I'm starting to see how the desired outcomes from an algorithm can encode biases.

If you are interested in learning more about the fake generated people heres a link to the website: https://this-person-does-not-exist.com/en and the paper https://arxiv.org/abs/1812.04948

The example of the twitter algorithm really struck me as interesting. When I thought of algorithmic bias, I mostly thought of big things like neural networks providing better home loans to zip codes and less about the mundane things.

Broader than race, it is extremely worrying that facial recognition systems could be used potentially in the future or even now to recognize "criminals." And especially with the lack of proper identification for many faces of color; the issue would be compounded.

I wonder if the Kuwaiti government would move as far as to take away citizenship from these Bidoon minorities if they submitted to a DNA test.

Really interesting how the official protocol recognizes that what they're doing could have legal and ethical issues which is why it can't be the only reason. Yet they continue to employ the practice anyways.

When the government is handing over private citizens genetic data to private companies this could be very problematic. Without the regulatory oversight needed these companies could sell data of citizens who never consented.

It is worrisome to see this dramatic juxtaposition between not having basic needs but having these UID and election cards that could be used to further prejudice people

This point is very powerful to me. Assuming the absurdity that genetic testing could possibly identify you were from one country, what's to stop your family from moving countries and subsequently being persecuted?

Seemingly in an endless cycle, IQ tests have been shown to have significant biases that are then reencoded into these "genetic markers for intelligence".

More recently, many prominent smart phone and camera manufacturers, i.e. Google, Samsung and Apple, have also made ads and showcased their ability to do something now that they should have done a long time ago.

This reminds me of the novel "Invisible Man" by Ralph Ellison which addresses many of the social issues black people faced and continue to face in the 1900s. The main character feels often socially invisible in this novel.

Interestingly, I have read that this false neutrality was present in both how film chemistries were developed and how digital photographic sensors were developed.