Case Study 8.1: Colorism and Dating App Preferences — What the Data Reveal

Background

In 2014, OkCupid co-founder Christian Rudder published Dataclysm, a book drawing on aggregated (anonymized) platform data from OkCupid's 25 million users. One of his most discussed chapters presented data on racial and skin-tone preferences in online dating: specifically, that message and response rates varied substantially by the racial identity of both sender and recipient, with consistent patterns that cut across gender lines.

The findings were widely reported — and widely contested. Understanding what the OkCupid data actually showed, what it did not show, and what it means for our understanding of colorism and racialized desirability requires careful analysis.

What the OkCupid Data Showed

Rudder's analysis documented that heterosexual women of all racial groups rated Black men's profiles lower on average than White men's profiles, and that heterosexual men of all racial groups (except Black men) rated Black women's profiles lower than White women's profiles. Asian men and Black women received the fewest matches and the lowest response rates; White men and Asian women received the most.

Critically, these patterns were not simply about racial group membership — they tracked within-group variation in skin tone as well. Lighter-skinned Black and Latina women received more messages and higher attractiveness ratings than darker-skinned Black and Latina women, even when other profile variables were held roughly constant.

The data also revealed that stated preferences did not always match revealed preferences: users who said they were "open to all races" in their profile settings showed racial preference patterns in their actual swiping behavior nearly identical to users who had selected specific racial preferences.

What the Data Cannot Tell Us

⚠️ Critical Caveat: The Limits of Platform Data

Several important limitations constrain what conclusions can be drawn from dating app data on racial preferences:

Self-selection. Users who join a particular platform are not representative of all people seeking partners. OkCupid's user base at the time was disproportionately young, urban, college-educated, and left-leaning — populations that may show different preference patterns than a general population.

Profile as representation. Attractiveness judgments on dating apps are based on photographs, bios, and platform context — all of which are shot through with racial signaling. A "Black woman's profile" is not just a Black woman; it is a curated self-presentation within a specific cultural context that carries connotations beyond individual appearance.

Preference vs. discrimination. There is ongoing debate about whether racial preferences in dating are analogous to other protected forms of discrimination. The philosophical question is genuinely contested; the empirical question of whether such preferences cause harm is less so.

The platform's role. Dating apps are not passive conduits for pre-existing preferences. Algorithmic ranking, the structure of the interface (swipe-based evaluation encourages rapid, appearance-focused judgment), and the social norms that develop within platforms all shape what preference patterns emerge and are reinforced.

Colorism Within Racial Groups

The more specific finding — that within the Black and Latina user groups, lighter skin tone was associated with more favorable outcomes — is particularly important for understanding colorism as distinct from inter-racial discrimination. This pattern is consistent with Margaret Hunter's sociological research on colorism (discussed in the chapter), and it suggests that dating app data is capturing not just racial preference but a finer-grained skin-tone hierarchy that operates within racial categories.

🔵 Ethical Lens: Do Platforms Amplify or Reflect Bias?

This is one of the most contested questions in research on dating apps and racial bias. Two positions are defensible:

The reflection argument: Platforms simply make visible preferences that pre-exist in the broader culture. The skin-tone hierarchy observed on OkCupid reflects the same hierarchy documented by Hunter and others in employment, education, and marriage markets. The app did not create it.

The amplification argument: The platform's design — rapid, photograph-based judgment without the context-building that characterizes offline attraction — strips away the information that might counteract implicit bias. When attraction is reduced to a photo swipe, demographic shortcuts become more powerful, not less. Algorithms that optimize for engagement may further concentrate attention on conventionally attractive (often lighter-skinned) profiles, creating feedback loops that intensify existing hierarchies.

The available research (Tyson et al., 2016; Hutson et al., 2018) suggests both processes are operating. Platforms do reflect pre-existing biases, and the design of specific platforms amplifies or attenuates those biases in measurable ways.

Discussion Questions

1. Should dating app companies be responsible for reducing racial and colorism-based preference patterns on their platforms? If so, what interventions would you propose? If not, why not?

2. Rudder's publication of the OkCupid preference data was controversial — some criticized it as legitimizing or naturalizing racial preferences. Was publication of this data ethical? What principles should govern the publication of preference data that reflects existing social hierarchies?

3. How does the gap between stated preferences ("open to all races") and revealed preferences (actual swipe behavior showing strong racial preferences) inform our understanding of how colorism operates? What does this tell us about the relationship between conscious values and implicit bias?