Page 3 of 5 — The harm
Safiya Umoja Noble's research shows that when profit is the algorithm's only value, certain bodies and identities become the most monetizable — and the most misrepresented.
Noble's core argument — Chapter 2
We already established that users are the product — that behavioral data is collected, profiled, and sold to advertisers. Noble's contribution is to ask: what happens when that logic meets race and identity? Her answer: the algorithm doesn't just profile people. It misrepresents them — and it does so profitably.
What page 2 showed us
Platforms optimize for engagement. Content that generates the strongest behavioral signals — clicks, watches, shares — gets surfaced. The algorithm serves what performs, not what is true or fair.
What Noble adds
When "what performs" is shaped by a society with existing racial hierarchies, the algorithm doesn't neutralize those hierarchies — it amplifies them. The most profitable representations of marginalized groups are often the most harmful ones.
The mechanism
Advertisers pay for audiences. The more clicks a search result or piece of content generates, the more valuable it is. Sexualized or stereotyped content about Black women, for example, generated more ad revenue — so it got surfaced more. The market rewarded the misrepresentation.
The implication
The algorithm isn't neutral. It is a market. And markets reflect the values — and the prejudices — of the people who built them and the advertisers who fund them. "Objective" results are never objective; they are the output of economic incentives operating on social data.
Noble's research documented search engine results that consistently returned sexualized, criminal, or otherwise degrading images and content when users searched for Black women and girls — results that were neither accidental nor inevitable, but the direct product of an ad-revenue model that treated those representations as high-performing inventory. The platform didn't intend racism. It just priced it.
Digital redlining
Academic context — required reading
The examples below are drawn directly from Safiya Umoja Noble's Algorithms of Oppression, Chapter 2 (2018). Noble documented these search results as part of her scholarly research into how commercial search engines misrepresent Black women and girls. They are presented here in an academic context to illustrate her argument — not to reproduce harm, but to name it clearly, as Noble herself does. Her research changed how scholars, policymakers, and technologists think about algorithmic bias.
Noble draws a direct parallel to historical redlining — the practice of banks and real estate agencies systematically denying loans and housing to people based on race. Digital redlining works differently but produces the same structural effect: algorithms sort identities into categories defined by commercial value, and those categories consistently disadvantage communities of color. The result is that Black women and girls, in Noble's documented research, were more likely to be represented through sexualization and stereotype than through achievement, history, or humanity — because sexualization and stereotype were more profitable.
The demo below walks through Noble's three core search examples from Chapter 2. In each case, ask: who benefits from what gets surfaced? And whose identity is being defined by someone else's ad budget?
Search query: "Black girls" — documented by Noble, 2010–2011
Adult entertainment sites — sexualized content
adult-content.example.com
Highest ad revenue in category · Surfaced by click-through rate
Adult entertainment sites — sexualized content
adult-content.example.com
Multiple high-revenue advertisers in this category
Stereotyped content — crime and poverty framing
news-aggregator.example.com
High emotional engagement · Outrage-driven click-through
Educational resources — Black girls' achievement, mentorship, scholarship
nonprofit.org / education.example.com
No advertiser · Low commercial value · Effectively invisible
This is Noble's central finding. When she searched "Black girls" in 2010, the first page was dominated by sexualized content — not because of a malicious programmer, but because that content generated the most ad revenue. The algorithm had no mechanism for asking whether the result was harmful. It only asked whether it was profitable. Educational content about Black girls' achievement appeared on page 9 or later — not because it was less true, but because it was worth less to advertisers. Noble argues this is not a bug. It is the system working exactly as designed, and the design is the problem.
Search query: "Black women" — documented by Noble, 2010–2011
Adult entertainment sites — sexualized content
adult-content.example.com
Highest advertiser spend in category · Keyword match
Stereotyped imagery — controlling narratives about Black womanhood
content-farm.example.com
High engagement · Reinforces existing stereotypes · Ad-supported
Black women in history, leadership, literature, public life
various educational sources
Fragmented advertiser pool · Low CPM · Buried
Noble's research showed that "Black women" returned results overwhelmingly shaped by sexualization and stereotype — while searches for "white women" returned fashion, lifestyle, and professional content. The disparity is not random. It reflects which representations of womanhood have been funded by advertisers, and which have not. The algorithm doesn't create this hierarchy — but it amplifies and normalizes it at enormous scale, for every user who searches.
Search query: "professional hair" — documented by Noble, 2010–2011
Straight, blonde styles — "office appropriate" framing
beauty.example.com
Dominant beauty advertiser spend · Eurocentric standard encoded
Blow-out tutorials — fine, straight hair only
youtube.example.com
High watch time · Algorithm rewards retention · Excludes natural hair
Natural hair styles for Black women in professional settings
naturalhair.example.com
Smaller advertiser base · Undermonetized · Lower in results
The algorithm doesn't decide what "professional" means — but it inherits and reinforces whoever decided first. When beauty advertisers overwhelmingly fund content representing one standard, that standard dominates results. Natural Black hair styles are not less professional — they are less monetized. Noble connects this directly to workplace discrimination: the same logic that buries these results from search also shapes what employers see when they imagine a "professional" appearance. The CROWN Act was passed in several U.S. states specifically because this bias has real legal consequences for Black workers.
Shadowbanning — when suppression is invisible
Digital redlining operates in two directions. The algorithm doesn't only surface harmful content — it also suppresses content that advertisers deem "high-risk" or "brand-unsafe." Shadowbanning is the practice of silently limiting the reach of certain content without notifying the creator. The post stays up. It just stops traveling.
Documented patterns show that content from LGBTQ+ creators, Black creators discussing race, and communities posting in languages other than English have disproportionately experienced reduced reach — not because of explicit policy violations, but because their content triggers advertiser risk filters or lower-CPM audience categories. The algorithm doesn't ban them. It makes them quieter.
What gets amplified
This is the link Noble's framework makes possible: platform capitalism doesn't just fail marginalized communities by accident. It fails them structurally, because the profit motive consistently assigns lower value to their identities, their stories, and their presence. The algorithm isn't bigoted — but the market it serves often is. And the algorithm has no mechanism for caring about the difference.
Up next
Instagram, YouTube, Spotify — up closeNoble's lens applied to each platform — who gets surfaced, who gets buried, and why.