#algorithms of oppression

Technology, simply defined, is information with a goal and a way to be reproduced. Writing is technology. So are telephone polls, hammers, knives, and AI. Technology has often been cited as the driving force of the United States empire. And for Western ways of living, technology has been seen as a gift to foreign regions. For the purposes of this essay, I will connect concepts of racism that have led to what we consider today as objective technology– that is, tech that is unbiased and bears science-backed ideas of truth. And thus, we will discover how the intersection of tribal sovereignty, (un)naturalized algorithms, and colorblind ideology all help us understand how technology has historically been used as a tool to boost colonial motives. Let’s first look at tribal sovereignty, what it means, and how it works in our modern world.

Tribal sovereignty, as defined by Marisa Elena Duarte, is the right for Indigenous and Native tribes to self-govern. But because white leaders have a penchant for manifest destiny, the Indigenous peoples of what is now known as “America” are forced into difficult situations like losing land and access to the internet. To understand how to move forward, Duarte asks us in her book “Network Sovereignty” to look at Indigenous spiritual practices and appreciation of the land in contrast to how colonial technology is developed under racist guidelines. She lists these considerations for the digital generation: colonial technologies have historically been weaponized against Native and Indigenous peoples, digital technology can support Native youth, and awareness of digital technology’s ability to limit traditional Native notions of grace and peace.

On that note, the “Natives in Tech” conference, which was held in 2020, has provided that there are big movements created by Indigenous and Native tribes to cut out third-party vendors and big tech companies and is moving forward with creating a network of Native software engineers. These software engineers move within the parameters of the sacred land connection integral to Indigenous ways of living. These movements teach us that it is possible to create systems of technology that fully encapsulate non-white and non-western existences and that doing so actually makes the spread of information found in the complexities of globalization.

Ever since the birth of search engines and the internet at large, we've been told that algorithms tell us more about ourselves than they do about the people who make them. What our screens show us is a product of what we’ve searched previously and what we’ve clicked on the most. The reality is algorithms are not naturalized. Algorithms are, actually, highly simplified systems based on the coder’s bias. Safiya Umoja Noble, author of “Algorithms of Oppression,” contends they currently “[do] not provide appropriate social, historical, and contextual meaning to already overracialized and hypersexualized people who materially suffer along multiple axes” (36). A big hint for why this is such an issue comes from the lack of Black, Indigenous, and People of Color employed in Silicon Valley. In “Silicon Valley Pretends That Algorithmic Bias Is Accidental,” Amber Hamilton discusses the tech culture, which has a history of racist and sexist hiring discrimination.

What’s more, Hamilton continues, is that tech companies, like Google, have a habit of dissuading employees from holding political discussions in the workplace. Yet again, we see an example of how easy it is for white tech to reflect white interests. We return, again, to the overarching idea that white tech seeks white power. Even if these instances seem unintentional, they tell us a story. The colorblind ideology that ensues, as Ruha Benjamin says in “Race After Technology,” “are sold as morally superior because they purport to rise above human bias” (38). It is almost impossible to challenge tech as we are brought to believe it's an entity all of its own, totally void of its creator's morality. MIT's data scientists work hard to construct robots without gender, class, or race. While the robots indeed were “servants” and “workers,” MIT scientists referred to them as “friends and children, addressing them in “class-avoidant” terms (Benjamin 42). Programmers felt so uncomfortable inputting the varying histories of racism, transphobia, and misogyny that they just let them out altogether. Unfortunately, acting as if these things didn’t exist doesn’t make technology better. It only makes it worse. So how do tribal sovereignty, naturalized algorithms, and colorblind ideology all tie together?

Image: Wangechi Mutu, 'You are my sunshine,' collage painting on paper, 24 x 36", 2015.

“Marginalized and oppressed people are linked to the status of their group and are less likely to be afforded individual status and insulation from the experiences of the groups with which they are identified. The political nature of [search engines] demonstrates how algorithms are a fundamental invention of computer scientists who are human beings–and code is a language full of meaning and applied in varying ways to different types of information” -Safiya Umoja Noble, "Algorithms of Oppression"

Before reading Safiya Umoja Noble’s “Algorithms of Oppression,” I was among the many that truly believed Google was a neutral public resource. Maybe you’ve heard the saying “library at your fingertips.”

However, despite popular belief, its algorithms serve paid advertising which happens to sport search results as a product of “most relevant and useful information” (Noble 37). Algorithms involve the process of “rendering,” a procedure that Noble calls “expressly social, economic, and human,” implemented through a series of steps called “algorithms” (37). The common misconception that Search is a mirror to human beliefs is combatted by Noble’s notion that “current algorithmic constraints [do] not provide appropriate social, historical, and contextual meaning to already overracialized and hypersexualized people who materially suffer along multiple axes” (36).

This means that for the white programmers who are uncomfortable talking about race, accurate depictions of Black, Indigenous, and People of Color escape the clutches of our everyday algorithms.

Furthermore, search engine optimization or SEO represents pushing ads or websites to the “top of a results list for a query,” providing profit to various companies or Google based on website clicks (Noble 40). Because Google is what Noble calls a “multinational advertising company,” it can prioritize search results for, say, “Black women” under a multitude of porn hyperlinks over the “eleven and a half billion documents that could have been indexed” (49).

Even someone such as Noble, who defines herself from the Black feminist perspective, a pedagogy that analyzes the intersection between racism, sexism, and so on, would find that search results do not always directly reflect her interests.

One quote that I would like to include, and that pulls away the smoke screen of “naturalized” algorithms, comes early on in the chapter, “Marginalized and oppressed people are linked to the status of their group and are less likely to be afforded individual status and insulation from the experiences of the groups with which they are identified. The political nature of [search engines] demonstrates how algorithms are a fundamental invention of computer scientists who are human beings–and code is a language full of meaning and applied in varying ways to different types of information” (Noble 26).

The master algorithm seems to give its name an inglorious connotation.

WEIRD comes to mind. White, Educated, Industrialized, Rich and Democratized. This problem of many humanities has also made it into the coding of algorithms. Previously in history, the problem was that many foundations for research used a too small and homogeneous pool of people and most of the study was done by white, male and wealthy people, so that their results were not representative for the whole population.

And that in two ways. On the one hand, prejudices, when they were not yet politically incorrect, flowed directly into pseudo-research. As emancipation and equality spread, it were only the indirect, unvoiced personal opinions. But the research leaders, professors and chief executives incorporated their conscious and unconscious worldviews into the design of questions, research subjects, experimental methods and so on.

Secondly, these foundations for the research were presented to an equally biased audience as well as test subjects. For a long time, much of the research has been based on these 2 foundations and is it indirectly until today, because new research builds on older results. It is like trying to finally remove a bug in the source code that evolved over the years with the bug as a central element of the whole system. That alone is not the only severe problem, but the thousands of ramifications it left behind in all other updates and versions. Like cancer, it has spread everywhere. This makes it very expensive to impossible to remove all these mistakes again. A revision would be very elaborate and associated with the resistance of many academics who assume dangers for their reputation or even whole work. They would see their field of research and their special areas under attack because nobody likes criticism and that would be a hard one to swallow.

What does this have to do with the algorithms? The programming of software is in the same hands as in the example above. Certainly not in such dimensions, but subconsciously inadvertent opinions may flow into it and the way search engines generate results is even more problematic. Especially in the last few years, deep learning, AIs, big data, and GANs, Generative Adversarial Networks, have been much more integrated into the development so that the old prejudices could begin evolving in the machines themselves without extra human influence.

This means that, in principle, no one can say anymore decidedly how the AIs come to their conclusions. The complexity is so high that even groups of specialists can only try timid approaches to reverse engineering. How precisely the AI has become racist, sexist or homophobic cannot be said anymore and worse, it can not be quickly repaired in hindsight. Because the reaction patterns on a search input cannot be selected in advance.

There is a sad explanation: Unfortunately, people are often infested with low instincts and false, destructive mentalities. When millions of people have been focusing their internet activity on aggressive hostility for decades, the algorithm learns to recognize their desires. There is a lot of money to earn and the AI should provide the users with what they want. The market forces determine the actions of the Internet giants and these give the public what it craves for. Ethics and individualized advertising can hardly follow the same goals. This is, even more, the case for news, media and publishers who suffer from the same problems as the algorithms. With the difference that they play irrelevant, rhetorical games to distract from the system-inherent dysfunctions.

The same problem exists with the automatic proposal function of online trade, which can inadvertently promote dangerous or aggressive behavior. Or with the spread of more extremist videos by showing new and similar ones, that are automatically proposed, allowing people to radicalize faster. The AI does its job, no matter what is searched for.

On a small scale, the dilemma has already been seen with language assistants and artificial intelligence, degenerating in free interaction with humans. For example, Microsoft's intelligent, self-learning chatbot, which was transformed into a hate-filled misanthrope by trolls within days. It is not difficult to imagine the dimension of the problem with the far-spread of new technologies.

One of the ways to repair these malfunctions is time. When people reset the AI's by doing neutral and regular searches. That's too optimistic, so it's more likely we will have to find a technical solution before people get more rational. For both better alternatives, the search queries and the results would have to change significantly before there could be the beginning of positive development.

The academic search results should not be underestimated. These often false, non-scientific foundations on which many of the established sciences stand. Even if the users became reasonable, there would still be millions of nonsensical results and literature. These fake, antiquated buildings of thought, on which many foundations of modern society are based, must be the primary objective. The results are the symptoms, but those dangerous and wrong thinkings are the disease. The long-unresolved history behind it with all its injustices has to be reappraised because it is the reason for widespread poverty and ignorance, which has its roots in wrong social models so that innocent AIs get deluded by search requests.

And the effects of search input and search results are mutually reinforcing. The mirror that they hold for society testifies only to hidden prejudices. However, those feel saver in their secret corners because they are supposedly unrecognized and subtly stoked by populists additionally and for their benefit. When wrong thinking has buried itself so deeply into a society, it also becomes part of all the products of that culture.

So you read So You Want to Talk About Race and now you have more questions. Specifically, you’re wondering how privilege affects your life online. Surely the Internet is the libertarian cyber-utopia we were all promised, right? It’s totally free of bias and discrimina—sorry, I can’t even write that with a straight face.

Of course the Internet is a flaming cesspool of racism and misogyny. We can’t have good things.

What Safiya Umoja Noble sets out to do in Algorithms of Oppression: How Search Engines Reinforce Racism is explore exactly what it is that Google and related companies are doing that does or does not reinforce discriminatory attitudes and perspectives in our society. Thanks to NetGalley and New York UP for the eARC (although the formatting was a bit messed up, argh). Noble eloquently lays out the argument for why technology, and in this case, the algorithms that determine what websites show up in your search results, is not a neutral force.

This is a topic that has interested me for quite some time. I took a Philosophy of the Internet course in university even—because I liked philosophy and I liked the Internet, so it seemed like a no-brainer. We are encouraged, especially those of us with white and/or male privilege, to view the Internet as this neutral, free, public space. But it’s not, really. It’s carved up by corporations. Think about how often you’re accessing the Internet mediated through a company: you read your email courtesy of Microsoft or Google or maybe Apple, and ditto for your device; your connection is controlled by an ISP, which is not a neutral player; the website you visit is perhaps owned by a corporation or serves ads from corporations trying to make money … this is a dirty, mucky pond we are playing around in, folks. The least we can do as a start is to recognize this.

Noble points out that the truly insidious perspective, however, is how we’ve normalized Google as this public search tool. It is a generic search term—just google it—and, yes, Google is my default search engine. I use it in Firefox, in Chrome, on my Android phone … I am really hooked into Google’s ecosystem—or should I say, it’s hooked into me. But Google’s search algorithms did not spring forth fully coded from the head of Zeus. They were designed (mostly by men), moderated (again by men), tweaked, on occasion, for the interests of the companies and shareholders who pay Google’s way. They can have biases. And that is the problem.

Noble, as a Black feminist and scholar, writes with a particular interest in how this affects Black women and girls. Her paradigm case is the search results she turned up, in 2010 and 2011, for “black girls”—mostly pornography or other sex-related hits, on the first page, for what should have been an innocuous term. Noble’s point is that the algorithms were influenced by society’s perceptions of black girls, but that in turn, our perceptions will be influenced by the results we see in search engines. It is a vicious cycle of racism, and it is no one person’s fault—there is no Chief Racist Officer at Google, cackling with glee as they rig the search results (James Damore got fired, remember). It’s a systemic problem and must therefore be addressed systemically, first by acknowledging it (see above) and now by acting on it.

It’s this last part that really makes Algorithms of Oppression a good read. I found parts of this book dry and somewhat repetitive. For example, Noble keeps returning to the “black girls” search example—returning to it is not a problem, mind you, but she keeps re-explaining it, as if we hadn’t already read the first chapter of the book. Aside from these stylistic quibbles, though, I love the message that she lays out here. She is not just trying to educating us about the perils of algorithms of oppression: she is advocating that we actively design algorithms with restorative and social justice frameworks in mind.

Let me say it louder for those in the back: there is no such thing as a neutral algorithm. If you read this book and walk away from it persuaded that we need to do better at designing so-called “objective” search algorithms, then you’ve read it wrong. Algorithms are products of human engineering, as much as science or medicine, and therefore they will always be biased. Hence, the question is not if the algorithm will be biased, but how can we bias it for the better? How can we put pressure on companies like Google to take responsibility for what their algorithms produce and ensure that they reflect the society we want, not the society we currently have? That’s what I took away from this book.

I’m having trouble critiquing or discussing more specific, salient parts of this book, simply because a lot of what Noble says is stuff I’ve already read, in slightly different ways, elsewhere—just because I’ve been reading and learning about this for a while. For a newcomer to this topic, I think this book is going to be an eye-opening boon. In particular, Noble just writes about it so well, and so clearly, and she has grounded her work in research and work of other feminists (and in particular, Black feminists). This book is so clearly a labour of academic love and research, built upon the work of other Black women, and that is something worth pointing out and celebrating. We shouldn’t point to books by Black women as if they are these rare unicorns, because Black women have always been here, writing science fiction and non-fiction, science and culture and prose and poetry, and it’s worthwhile considering why we aren’t constantly aware of this fact.