Are Search Engines Blackwashing Their Search Results?

10 points by affiliator ↗ HN
google "portraits of European people":

https://www.google.com/search?q=portraits+of+European+people&source=lnms&tbm=isch

Here are my results:

http://i.imgur.com/6s9qK9g.jpg

Similar results with bing, duckduckgo.

No so with Yandex:

http://i.imgur.com/zvbwMuF.jpg

An Alt-Right vlogger has done a video on the phenomenon. The same thing happens with the search term "European People Art"

https://www.youtube.com/watch?v=b14IgSPkPC4

11 comments

[ 1.4 ms ] story [ 36.0 ms ] thread
(comment deleted)
(comment deleted)
No, they're not, and if you look at what pictures are actually being link you can start to understand it a bit more.

What the first result is returning is keyworded based on art discussions about European works. Relevant articles discussing the uniqueness of portraits of black people in rennaissance paintings are the results you're seeing. Likely it's the inclusion of "people" which is throwing the results, as you see a much "whiter" result set if you search "portraits of european" (plural or not doesn't matter). In fact, for the original search if you scroll down you begin to see the paintings and portraits you are expecting to see.

Basically, it's just keywording providing results you weren't expecting. The word that throws it is "people".

Edit: and in regards to why Yandex isn't showing it, it's because the top hits are from a russian artist's site, so they're just weighting russian sites higher. For whatever reason, people clicked to these images when they searched it on Yandex.

This is a non-story.

There is a difference in the results for "portraits of European" and "portraits of Europeans".

"portraits of Europeans" contains a large number of non-European faces.

> "portraits of Europeans" contains a large number of non-European faces.

Which are the same that are showing up via the articles about art of european history from the original result. Again, this is just keywording working against your expectations. The relevant results for these keywords lead to articles discussing the presence of non-whites in European artwork.

What you envision for the results of the search terms are apparently different than how other people envision them. In this case, the search is considering portrait paintings, not portrait photos or anything, and from there, it looks like the most popular results are these articles.

Search for "[pictures, photos] of european people" or just "european people" and you get more pictures of white europeans. Unsuprisingly, once you search "paintings of european people" you start to get the aforementioned paintings once more.

This isn't a conspiracy or social manipulation, it's just that there is apparently enough discussion and interest in older european paintings including black people that searches related to paintings of europeans returns a few results with black people or indian people.

so you are think websites containing the keywords 'portraits', 'Europeans', and 'people' happen to predominantly be talking about portraits of non-European people by Europeans?
Yes, and the search results show that there are. I'm not sure why you're even questioning this and not just looking at the search results beyond the color of the people in the paintings:

http://medievalpoc.tumblr.com/post/62747229987/marten-van-my...

http://afroeurope.blogspot.ru/2010/08/history-of-black-peopl...

http://realhistoryww.com/world_history/ancient/Misc/Art/Addi...

Yes, blogs and tumblr sites but these are valid results.

Well if they are blackwashing the image results, they would be blackwashing the regular results too. I am not 100% convinced but I think you're best argument is that a search for "european people" gives the results you would expect. If they were blackwashing results, you would expect to see the same phenomenon there.

What do you make of the results for the search term "American inventors" which returns largely black inventors.

http://i.imgur.com/uTF3FTN.jpg

I think it reflects the reality of what people apparently consider great: https://en.wikipedia.org/wiki/Category:American_inventors

Though, admittedly after Edison, Carver, Whitney, Franklin, Moorse, I ran out of people I consider "famous" inventors, so the list doesn't really impact me. I do question their inclusions for some inventors though...as quite a few were not born in the US, and only came to the US later on, but that's my own hang up)

I also suggest that this is a prepared result and not a search result so I don't hold it to the same scrutiny. I treat it the same as an advertisement.

There might be a story here; a story about a company that tinkered too much with their search algorithm. I don't know what they've been changing over there, but their search seems to get worse and worse every year. If the application of AI to my search results means any unusual words are weighted out of the equation, I'd rather just have the "dumb" results.
This might make an interesting blog post, particularly if supporting data was exposed. Applying the Russell-Norvig definition of an artificial intelligence as an actor capable of action in the world based upon its perceptions, then what we may have is the manifestation of narrow perception.