I want to say right at the beginning of this post, that our partnership with Google is extremely valuable and that they have been invariably helpful and generous toward us.
I also want to say that this is not just a Google problem. This seems to be a problem with algorithms that support predictive search. They’re an unintended result of people trying to make our internet searching for service, content and information faster and more successful.
I also need to tell you that when I brought this matter to attention of Google, I got a same day response along with the promise to try and see what was going wrong and to “fix the issue”.
So where did all this come from?
Last week, my attention was grabbed by the concern that rippled around the world, about the differential in results when one Google searches “3 Black Teenagers v 3 White Teenagers”
As a concerned New Zealand Educator who campaigns for equity, access and success for Maori & Pasifika, I decided to conduct the same test but change the terms. I selected “Maori" as the first term and decided not to use “White” as the second term as it would be likely to return the same result as the viral one, so changed that term to “Pakeha” (the Maori word for White or European)
So... my 1st search was <3 Maori Teenagers>, my 2nd search was <3 Pakeha Teenagers>
I used Google Images as the input medium.
Here are the results:
I’m very disconcerted by the results of the search.
As the search did with the original terms (White v Black) the return showed mostly less than salutary images of Maori.
Just like the <White> search the <Pakeha> search returned groups of well dressed kids, in this case all in school uniform and in each case, in front of schools. Some of the kids in the <Pakeha> search are in fact Maori and Pasifika, but are portrayed respectful, beautifully dressed and in front of schools as Pakeha.
Here’s what I’m so disconcerted about:
Just like Pakeha, Maori kids go to school. In our country, unlike the US, Maori kids wear uniform just like Pakeha kids.
How is this weird differentiation occurring, that portrays Maori as not wearing uniform, not associated with school, when we work in and love a community where Maori kids go to school and all wear uniform.
The search could even have returned Maori kids in Kapa Haka uniforms, Rugby or Netball uniforms, or doing waka ama.
This is seriously weird.
I’ve been told that Google searches are informed by the typical content interests of the searcher over time.
Not sure if this is the case with Images, but what jangled my nerves even more is that my account <email@example.com> is associated directly with Pt. England School, where Maori kids wear uniform and go to school! Even more surprising ptengland.school.nz is a Google Apps for Education Domain as is manaiakalani.org. As most of the delightful learners associated with these domains are Maori and Pasifika, one would expect that the predictive search, knowing who I am (and on my phone, where I am) would dish me up some top 10 images that showed me some of what I expected. Unfortunately what the search showed me was what the media dishes up at 6:00 p.m. in its attempt to get ratings.
It appears to be the word <teenager> that dominates the search expectation, as though to be Maori and a teen has an inbuilt negative expectation.
Here is the search result of <Maori Pt England>
it generates a very different picture
I’ve become very interested and concerned about what the juxtaposition of the word <Maori> with the word <Teenager> does.
It seems to do something very wrong
I concluded this informal research by searching Images under the search term <Maori Tamaki College) as Tamaki is our Manaiakalani High School
The return was as I hoped for
Here’s the thing: What you see above are Maori Teenagers!
I know I’ve laboured the point, but I feel it’s important
I have no idea what can be done about it, but I hope something can as I know this is not what Google intends at all and I’m sure it poses a problem for other companies and developers who are refining predictive search as part of what they bring to market place.
Of course, the implications of the issue I've raised here, go far beyond this simple test associating ethnicity with age bracket. This has implications for what we and our learners have presented to us when we search about religion, politics, and contemporary issues to name just a few.
We need to know that along with freedom of speech, we are experiencing freedom of access to information, service and content where the information set is filtered by us not controlled by others.
Thanks for listening/looking
We do still love our partnership with Google!