Is this bad? Should they have done something different? I don't know. I guess it comes down to what you expect when you search for those terms, who is searching for those terms, and if these lists have proven to be helpful to those searching for these terms.
As someone else already insightfully said results for American Scientists results a result set an awful lot like African American Scientists whereas US scientists or United States Scientists returns a results set more like you would expect.
In 2015 google photos was classifying black people as Gorillas that wasn't on purpose either.
https://www.theverge.com/2018/1/12/16882408/google-racist-go...
Trump is just living it? Why did you add that part? I pointed out that the results looked odd. When I opened up the comment section, every other comment was saying nothing was odd. Those comments you're referring to didn't exist yet. Now that I've seen them, yes, they do bring up an interesting point.
I'm not a Trump supporter. I'm not a republican. And nothing I said indicated that Google was out to get conservatives. There was no reason to bring up the Google image story or say anything about Trump in your reply.
"Google Attacked by Trump With Claim It Is Burying Conservative News" I was referring to the subject of the entire discussion not you or your politics. I was trying to be clever by replying to your statement AND the topic simultaneously.
Google images misclassifying black people as gorillas, as well as google instant answers providing erroneous and prejudicial information are examples of situations in which we could discover meaning where none exists in data generated by an algorithm.
It seems vastly more likely that the words american inventors appears many times in documents that begin african american inventors than google specifically tweaking that query to increase diversity the same way I don't think google meant to convey that black people were like gorillas or that Obama was king of America
OTOH, I'll bet there are websites that talk about, say, "George Washington Carver, an American scientist you have never heard about."
That said, the quality of curation seems rather poor.
Oddly, when your search isn't a substring of "African American Inventors" you don't get weird results.
It's almost like the search algo isn't perfect.
The original article claims something similar albeit on a different search result.
the underlying idea is that google is not just using pagerank and algorithms to show us the current state of the internet. But rather is adding their own bias to push up / down certain results.
That is unfair. This result is very different from a search like "USA inventors" which is semantically similar and it is an interesting discussion to have, why that is. I make no claims about what the results 'should' look like from a societal perspective.
And ultimately I agree with your criticism, it was too hasty on my part to claim google did this purposefully. Some of the ideas in the sibling threads offer very plausible alternative reasons which could have caused this difference.
One thing I've wondered with stuff like this is why we assume Google is the root cause here. Mind you, I've switched all my products away from Google, so I'm not defending them, BUT;
Reddit users frequently take part in manipulating Google search results for the sakes of memes. To put someones picture up on a specific search, for example. Yet other search results we believe to be hand crafted by Google.
Now, I think it is more likely that Google is manipulating data here. However, I never see it discussed why we don't think someone else could be manipulating it?
Furthermore, to be honest, this “organic” search has largely been used to debunk racist claims. And the search has been used for over a decade.
Shocking.
If you lay out the game theory brining this up in this context you are implying that this is part of some vast liberal conspiracy. That is the problem.
If the article is on Trump claiming that Google politically biases search results, Google having a history of returning suboptimal results in a way that is "stereotypically liberal" would be some decent supporting evidence - someone having done something before increases the prior probability of them having done so again. If that's actually accurate, of course.
Now, it seems pretty clear to me that this was a bug (and one I see in other, non-political search results) and that lowers the credibility of the criticism. I'd still rather have both sides make their strongest arguments though, and that includes pointing out prior occurrences that might suggest a pattern.
I don't think there is any overrepresentation in the result set, though there does seem to be a small cluster at the top (the top 5 and 9 of the first 12, but then a sudden drought.)
If you Google search "African American Scientists" and "African American inventors", you get almost the exact same photo list as "American Scientists" and "American inventors".
George Washinton Carver, Elijah McCoy, Madam CJ Walker, Lewis Howard Latimer, and Garrett Morgan
How many of those do you suppose are actually in the "Top 10" of US inventors
Thomas Edison, Alexander Graham Bell, Samuel Morse, Eli Whitney, Nikola Tesla, George Washington Carver, Charles Goodyear, Lewis Howard Latimer, Henry Ford, George Eastman
... which does corroborate OP's "American inventors" hypothesis.
Secondly google search results are a best guess based on prior search history and other data about you about what you want to read right now given a term.
You wont anywhere find a universal fact about what google themselves thinks about anything without reading material written BY google and not generated results from its search engine.
For me, "there are literally zero black", was not true. predominately not black, but Washington was in the list. I don't mean to be pedantic, but I take 'literally' seriously
Wherein they try to provide actual answers to questions more complicated than 2+2 or what is the weather like they have ended up with interesting results drawn from the nonsense that is the internet.
"Instant answers have attracted criticism for years. In 2014, Google results highlighted a Breitbart story saying Barack Obama was the king of America, and the next year, Google got flak for providing an excerpt from a creationist website saying dinosaurs were tools of indoctrination in response to the question, "What happened to dinosaurs?"
Conspiracy-mongering instant answers drew attention in March after the website Search Engine Land pointed out that Googling "Is Obama planning a coup?" displayed an answer from a website called SECRETS OF THE FED saying Obama was indeed planning a coup with the help of "the communist Chinese."
https://www.sfgate.com/technology/article/google-instant-ans...
Imputing bias from search results may be a mistake. Especially when in some cases it might even be a correct guess at what the searcher actually wanted to read.