The article mentions that queries made in the country's native language tended to be less "racy", suggesting the later. More generally, it's the differences in the intent of the search that may be the cause, with those intents reflecting the beliefs of others about a country.
> intents reflecting the beliefs of others about a country
Or simply the different intent of foreigners vs locals. If you search for "mujeres americanas" you're probably not interested in politicians or women in culture and media- you want to see the stereotypical "american woman". If you're based in the US and search for "american women" though you know the stereotype enough and you're most probably looking for different.
There is no correct, unbiased way to rank images (or anything). There may be a "homogenizer" which reshuffles the ranking based on image features but you would have to introduce an artificial category, e.g. nationality.
But I doubt a "de-bias-reshuffler" is a clearly defined "thing" let alone something that is easy to explain to the algorithm-scared public.
Since there is no "correct" way to rank images, why doesn't Google return them randomly? Because they have _chosen_ a way to rank them, motivated by some goal (likely, "user satisfaction"). It is therefore reasonable to critique that choice.
The article has cause and effect reversed. Peoples stereotypes, expectations, and "coded" language has created the results. When someone googles "Brazilian Women" or "Latina Women" they looking for porn and racy photos not pictures of normal women from those countries/regions.
There is all sorts of coded language that people use to get the results you want. Search "Black Women" vs "Ebony Women" and look at the difference in results. No one is searching Ebony unless they are looking for racy content.
> "It is no surprise that search engines replicate biases that are not exclusive to the technology, but rather cultural ones. Women of certain nationalities are pigeonholed into sexual and services roles by a male English-speaking culture," she adds.
What an absurd article. First, it offers no evidence that the search results influence society. I mean, does anyone look to Google image search to tell them how to perceive women? Do Google image search results influence perceptions at all? Maybe, but the article fails to make that point at all, other than in the title where it claims the results “cement” perception.
Second, "If an algorithm is not doing anything to compensate for this, then it'll be mostly racist, sexist, patriarchal.” That is, if an algorithm is purely a mirror reflecting a racist, sexist, patriarchal society, the algorithm is racist, sexist, and patriarchal. I contend that “if you’re not with us you’re against us” narrative is dangerous. Political winds can shift quickly. Arguing that neutrality is complicity will backfire as soon as the winds shift against you.
I whole heartedly support fighting he biases reflected by technology. I believe there are far more effective ways than the fraught minefield of building in bias correction.
9 comments
[ 4.0 ms ] story [ 27.6 ms ] threadOr simply the different intent of foreigners vs locals. If you search for "mujeres americanas" you're probably not interested in politicians or women in culture and media- you want to see the stereotypical "american woman". If you're based in the US and search for "american women" though you know the stereotype enough and you're most probably looking for different.
There is no correct, unbiased way to rank images (or anything). There may be a "homogenizer" which reshuffles the ranking based on image features but you would have to introduce an artificial category, e.g. nationality.
But I doubt a "de-bias-reshuffler" is a clearly defined "thing" let alone something that is easy to explain to the algorithm-scared public.
There is all sorts of coded language that people use to get the results you want. Search "Black Women" vs "Ebony Women" and look at the difference in results. No one is searching Ebony unless they are looking for racy content.
> "It is no surprise that search engines replicate biases that are not exclusive to the technology, but rather cultural ones. Women of certain nationalities are pigeonholed into sexual and services roles by a male English-speaking culture," she adds.
Second, "If an algorithm is not doing anything to compensate for this, then it'll be mostly racist, sexist, patriarchal.” That is, if an algorithm is purely a mirror reflecting a racist, sexist, patriarchal society, the algorithm is racist, sexist, and patriarchal. I contend that “if you’re not with us you’re against us” narrative is dangerous. Political winds can shift quickly. Arguing that neutrality is complicity will backfire as soon as the winds shift against you.
I whole heartedly support fighting he biases reflected by technology. I believe there are far more effective ways than the fraught minefield of building in bias correction.