Faces of the Internet: Representation Activity

The “Representation” Activity required us to do the following:

Perform Google image searches for the following terms

Teacher: almost exclusively white women

Professor: almost exclusively white men

Doctor: almost exclusively white men

Nurse: almost exclusively white females

Baby: almost exclusively white

Teenager: almost exclusively white, mainly girls

Criminal: Mostly clip art. When I did a search of faces only, they were disproportionately black, with some individuals who were white, and all men

What is striking about the results? Are you surprised?

What’s striking is how clearly every single search resulted in a “type” of either race or gender, and usually one involving both. Furthermore, the race was always white. I expected some bias in the results, but I expected it to mostly be limited to the searches like “criminal” and “doctor” for race and “nurse” and “doctor” for gender due to real-world differences in who usually holds these titles. But it makes no sense for “baby” to result in mostly white babies.

How are those results “chosen”? Are they meaningful at all? If we wanted to alter the results, what would it take?

It seems to me that the results are chosen based on the most prototypical example of each term, basically meaning whatever people tend to be looking the most often when they search those terms. Based on the fact that America is a nation composed of a white majority, this could explain why the the most typical example of any of these categories results in mostly white individuals. One piece of evidence that suggests this to me is the notable absence of red-headed, freckled, or green-eyed people, all of which are sizable minorities but considered atypical. If this is the case, to alter these results, an effort must be made toward adding diversity in the results and consciously correcting for our tendency to view white the default for everything. Another credible explanation is that the lack of women and racial minorities in fields of technology may explain these results. I read Al Jazeera’s piece on the subject, which discussed an effort to remedy this problem by having users seek to change what appears for a Google Images search of hands. It doesn’t seem to have worked.

Then choose other terms that you think might show the same kinds of patterns in their search results, and search for them. Were your predictions correct? Why (or why not)?

Dead: Interestingly, this turned up one of the more diverse sets of results. I finally saw one Asian face, as well as seeing a mix of men and white women.

Tired: Turned up males and females of various ages, but all white.

Female doctor: I searched this wondering if eliminating males would result in more diverse results. Well I was disappointed–it was still almost exclusively white.

Male nurse: Relatively diverse results.

And on a happy note,

Beautiful: Very diverse results including multiple minority groups.

My predictions were mostly correct. What surprised me is that even when there were black people represented, it was mostly men as far as I remember. Scott Westerfeld’s novel So Yesterday actually makes note of this phenomenon, dubbing it the “missing black woman trio” phenomenon. I thought it was just an exaggeration, but now I’m convinced. Most strikingly, as I predicted based on the results of the previous set of searches, there are almost no Asians, Hispanics, or Native Americans that stand out from the results. Suddenly I understand the “issues” component of this course.

Overall, it’s a sobering lesson on how much white really seems to be the norm on the Internet. But it’s nothing new. For instance, the technology of color film also once had a racist history.

Hopefully the Internet can grow more diverse in the future, just as film eventually did. Until then, we need to be mindful of this reality.

Leave a Reply

Your email address will not be published. Required fields are marked *