Abstract
Our days hardly ever go by anymore without ‘googling’ – the weather, arrival/departure times, opening hours of local businesses, meanings of some rare or complicated words, and a myriad other topics. The convenience of a one-stop search discourages most of us from questioning the results delivered onto our devices. We tend to take them for accurate, objective and neutral.
Safiya Umoja Noble’s book Algorithms of Oppression: How Search Engines Reinforce Racism disabuses us of this notion and reveals that in the absence of public debate about them, algorithmically induced decisions tend to ‘reinforce oppressive social relationships’ (p. 1) that further disempower the most vulnerable groups in society. Through its six chapters, the book demonstrates negative effects of algorithmic culture, paying particular attention to ‘the commercial co-optation of Black identities, experiences, and communities’ (p. 2).
Chapter 1 diagnoses the prevalence of business interests over social justice in the operating logic of tech companies that exploit racism and sexism as profitable attitudes. Substantiating the case for the search engine’s sexism, Chapter 2 reports on the query ‘black girls’ and notes that other groups of people (teenagers, doctors, professors, etc.) risk becoming similarly commodified and objectified if commercial platforms control the narrative about them (pp. 94–109). Chapter 3 addresses digital platforms’ ‘highly optimized’ content (p. 116) in order to keep viewers ‘on screen’ and to sell their attention to advertisers, just like TV channels do. Chapter 4 presents several highly publicized cases such as a search for ‘black on white crime’ (p. 121) that spurred to action Dylann Roof who shot dead nine African Americans in their church in Charleston, South Carolina, or the disappearance of small businesses as their word-of-month advertising is being undermined by Yelp searches. Here, the author also discusses tech companies’ global strategies of dodging local regulations as, for instance, in storing sensitive information on servers abroad and, by doing so, bypassing national regulations regarding personal data storage. Chapter 5 turns to the question of knowledge in the world where Google is a preferred source of information and where the ontologies it inherited from traditional classification systems preserve certain blinders and preclude search engines from functioning as ‘windows’ on the world (Alex Galloway as cited on p. 148). In closing, Chapter 6 contemplates the future of information culture that heavily depends on the web and calls for action in order to adopt public policy that would protect the collective rights of online users.
Noble’s motivation to investigate how search engines are implicated in the system of domination stems from her personal experience, namely, from encountering a barrage of pornographic images returned for her ‘black girls’ search query. While she momentarily entertains the idea that racism and sexism in the top search results might reflect ‘willful neglect’ (p. 5), she finds a profit imperative to be a more plausible explanation since corporations exist to maximize profit, not public good, she contends.
Notably, bias plagues not only Google’s searches and autocompletions. More established and authoritative classification systems such as the Library of Congress Subject Headings or Dewey Decimal Classification (hardly profit-making or profit-focused entities) also have to comb their indices from time to time to correct for racist, sexist, homophobic or otherwise offensive terms, reminding us that the systems developed by the powerful to organize knowledge and to buttress their interests continue doing so. Consequently, Noble refutes the argument that a search simply returns and magnifies covert stereotypes and prejudices of those who type in the search box. Instead, she charges search engines with racism being part of ‘the architecture and language of technology’ (p. 9). Unfortunately, the author does not theorize either the process or the mechanism that maps oppression onto digital spaces aside from confirming bias inherited from the existing classifications. Moreover, as she describes alternative arrangements deployed by tech companies in the European Union (EU) to allow more individual control over personal data and to recognize the ‘right to be forgotten’, she becomes convinced that Google – the only tech company whose searches she examines – is being uncooperative in adjusting its algorithms. To remedy the situation, the author proposes a legal protection delivered via policy regulation.
There are a few important omissions in the narrative laid out by the book. First, the author notes that online search is ‘different from traditional offline information seeking behavior’ (p. 37). Yet, she does not name those differences nor does she explain how search works. Second, even though Google’s famed PageRank is a proprietary algorithm unmatched by the breadth and depth of its reach, by the time Noble’s anti-Google crusade started, other search engines had been built and implemented. Do they render similarly biased results? If not, we need to qualify the book’s central proposition about racism being baked into algorithms. Third, asking ‘who decides’ (p. 18) on how search results get arranged/ranked, Noble’s probe remains highly abstract and renders highly abstract, probable culprits in neoliberalism and corporate greed. Apparently, the author did not interview search engine engineers who write code and whose ‘egregious tech designs’ (p. 70) affect marginalized groups in such a negative way. Such methodological choices amount to the study’s missed opportunity to reveal how algorithms are increasingly becoming integral in decision- and meaning-making in many spheres of contemporary life – from dating to insurance to national foreign policy. Having demonstrated that racist and sexist bias in search results is not an accidental aberration, the author overlooks other features and functions of search engines that engineers strive to model, for instance, relevance, a multilingual return, a search across diverse formats (videos, images and text documents), or the efficiency and expediency of a search that goes over several million pages in a split second, thus creating expectations in users about accurate, relevant and exhaustive information being readily available.
Despite these shortcomings and despite the acknowledged obsolescence of ‘data’ in this study, it offers valuable lessons worth hearing and repeating. First, it reminds us that any information organization and categorization is value-laden and that critical thinking skills are ever more important for navigating the digital as well as analogue world. Second, it highlights the materiality of the digital, the virtual and the online. Behind those novel spaces are infrastructure and human labour. Their materiality bleeds into the digital, and the tangled relationships in which they are embedded condition the resulting digital output. While the book raises more questions than it attempts to answer, its highly accessible narrative makes it suitable both for the general public interested in the subject and for those academics who seek to answer the burning questions of the day.