🔬 physics

Interplay between social contact and media exposure in the overestimation of racial diversity in the U.S

This study reveals that Americans systematically overestimate the size of racial minority groups across all geographical scales, with white respondents' local misperceptions driven by direct social contact and national ones by news coverage, while frequent social media use exacerbates these overestimations, collectively fostering an "illusion of diversity" that may undermine support for equity policies.

Original authors: Clara Eminente, Henrik Olsson, Ljubica Nedelkoska, Rafael Prieto-Curiel, Mirta Galesic, Elisa Omodei

Published 2026-03-31

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Clara Eminente, Henrik Olsson, Ljubica Nedelkoska, Rafael Prieto-Curiel, Mirta Galesic, Elisa Omodei

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to guess how many red marbles are in a giant jar. You can't see the whole jar, so you have to make a guess based on what you know.

This paper is about a very strange game Americans play: guessing how many "People of Color" (PoC) live in their neighborhood, city, state, and the whole country. The researchers found that almost everyone guesses wrong, and they guess that there are way more minority people than there actually are.

Here is the story of their findings, broken down with some simple analogies.

1. The "Zoom Lens" Effect

The researchers asked people to guess the numbers at four different "zoom levels":

Neighborhood: Your immediate block.
City: Your town.
State: Your entire state.
Nation: The whole United States.

The Finding: The further you zoom out, the bigger the mistake gets.

At the neighborhood level: People are actually pretty close to the truth. If you live in a diverse block, you see diversity. If you live in a white block, you see that.
At the national level: Everyone goes wild. People think the country is much more diverse than it really is. It's like looking at a tiny speck of dust through a microscope and thinking the whole room is filled with glitter.

2. The "Mirror" vs. The "TV Screen"

Why do we get it wrong? The researchers found that the answer depends on how far away the group is from you.

For your Neighborhood (The Mirror): You rely on direct contact. If your friends, family, and coworkers are mostly white, you think your neighborhood is mostly white. If you have many friends of color, you think your neighborhood is diverse. Your social circle acts like a mirror; you see what's right in front of your face.
For the Nation (The TV Screen): You can't know everyone in the US personally. So, you switch to indirect contact: the news and social media. This is where the distortion happens.

3. The "Echo Chamber" of Social Media

This is the most surprising part of the study. The researchers looked at how people get their information:

Reading the News (The Librarian): People who read the news frequently were actually better at guessing the numbers. The news tends to give facts and statistics, which helps correct the "diversity illusion."
Scrolling Social Media (The Wild Party): People who spend a lot of time on social media (like X, Instagram, TikTok) were the worst at guessing. They thought there were way more minority people than there actually were.

The Analogy: Imagine a party.

Traditional News is like a host who hands you a guest list with the real numbers.
Social Media is like a loud, chaotic dance floor where a few people are shouting about how crowded it is. Because social media mixes "news" with "personal stories," it creates a feeling that the room is packed with people of color, even if the guest list says otherwise. It creates an "Illusion of Diversity."

4. The "Own Group" Bias

The study also found something interesting about who is making the biggest mistakes:

White people tend to overestimate the size of minority groups, but they are "okay" at it.
People of Color actually overestimate the size of their own group even more than white people do!

Why? If you are part of a minority group, you might feel like you are everywhere because you are hyper-aware of your own identity. It's like wearing a bright red hat; you notice every other red hat in the crowd, making it feel like there are thousands of them, even if there are only a few.

5. Why Does This Matter? (The "Mission Accomplished" Trap)

The paper ends with a warning. This "Illusion of Diversity" is dangerous.

If people think, "Wow, the country is already so diverse! We have so many people of color everywhere," they might stop supporting policies that actually help create fairness and equality. They might think, "We've already won the game!" when, in reality, the game is still being played.

Summary

We all overestimate diversity, especially when thinking about the whole country.
Local guesses are based on who you know (your friends).
National guesses are based on what you see online (news and social media).
Social media makes the problem worse by making diversity look huge and everywhere.
The Danger: This false belief makes people think we don't need to fix inequality anymore, because they think we are already "fixed."

The researchers are basically saying: "Don't let the algorithm trick you into thinking the world looks different than it actually does."

1. Problem Statement

The study addresses the systematic phenomenon where the general population overestimates the size of racial minority groups (specifically People of Color, or PoC). While prior research has established this bias at the national level, significant gaps remain regarding:

Geographical Scale: How misperceptions vary across nested scales (neighborhood, city, state, nation).
Group Differences: How overestimation patterns differ between White respondents and PoC respondents (who often overestimate their own group size).
Mechanisms: The specific roles of direct exposure (social contact) versus indirect exposure (news and social media) in driving these misperceptions.
Consequences: The "illusion of diversity," where overestimation leads to the erroneous belief that representation goals are already met, potentially undermining support for equity-promoting policies.

2. Methodology

Data Collection

Survey Design: A purpose-built online survey administered via Prolific between May and June 2024.
Sample: $N = 866$ U.S. residents (after cleaning 134 invalid responses from an initial 1,000). The sample is representative of the U.S. population by gender, age, race, and geographical distribution.
Task: Respondents estimated the percentage of PoC in four nested geographical levels:
1. Social Circle (family, friends, colleagues, acquaintances).
2. Neighborhood (based on zip code).
3. City/Town.
4. State.
5. Nation (USA).
Ground Truth: Estimates were compared against 2024 U.S. Census Bureau data (American Community Survey, Table B03002) at the corresponding geographical resolutions.
Variables Measured:
- Demographics: Race, income, education, age, political affiliation.
- Exposure: Composition of social circles (percentage of PoC contacts), frequency of news consumption, frequency of social media use, perceived frequency of PoC coverage in news, and perceived tone of that coverage (positive/negative).
- Confidence: Self-reported confidence in estimates (1–6 scale).

Analytical Approach

Bias Classification: A response was classified as an overestimation if the respondent's estimate exceeded the actual census value by more than 10% ( $\beta$ ).
Statistical Testing:
- Bootstrapping: 1,000 bootstrapped samples were generated to determine the distribution of overestimation rates ( $f_{over}$ ) for White and PoC groups at each scale.
- Hypothesis Testing: Two-sided Mann-Whitney U tests assessed statistical significance between groups.
- Effect Size: Cliff's delta ( $\delta$ ) was used to quantify the magnitude of differences between distributions.
Machine Learning (Predictive Modeling):
- Model: Eight Binary Random Forest classifiers were trained (2 racial groups $\times$ 4 geographical levels).
- Goal: Predict whether a respondent is an overestimator based on exposure variables and sociodemographics.
- Evaluation: Precision-Recall Area Under the Curve (PR-AUC) was used to handle class imbalance.
- Interpretability: SHAP (SHapley Additive exPlanations) values were calculated to determine feature importance and the direction of influence (positive/negative impact on overestimation probability).

3. Key Results

A. Scale and Group Dependence

Scale Effect: The probability of overestimation increases progressively from local to national levels. At the national level, >50% of both White and PoC respondents overestimate PoC population size. At local levels (neighborhood/city), overestimation rates are lower (though still >40%).
Group Effect: PoC respondents overestimate their own group size more frequently than White respondents do, particularly at the neighborhood and national levels.
Implication: As the geographical scale expands, reliance shifts from direct observation (experience-based) to indirect information sources.

B. Mechanisms of Exposure

Direct Exposure (Social Circles):
- White Respondents: At local levels (neighborhood/city), the composition of social circles is the primary correlate of overestimation. Those with more PoC contacts overestimate more. This effect diminishes significantly at the state and national levels.
- PoC Respondents: Those with no White contacts in their social circle are more prone to overestimation at city and national levels compared to those with mixed circles.
Indirect Exposure (News & Media):
- News Frequency: Higher frequency of news consumption is associated with reduced overestimation for both groups.
- Social Media: Higher frequency of social media use is associated with increased overestimation for both groups.
- News Tone:
  - Local Level: Positive news tone correlates with higher overestimation (likely reinforcing salience of direct contact).
  - National Level: Negative news tone correlates with higher overestimation (consistent with "hyper-visibility" of minorities in crime/negative stories).
- Coverage Frequency: Perceived high frequency of PoC coverage in news correlates with higher overestimation, especially at the national level.

C. Integrated Model Findings (Random Forest)

Variable Importance: SHAP analysis confirms that exposure variables (social circle composition, news coverage frequency, news tone) are among the most influential predictors.
Scale Shift:
- At local levels, social circle composition is a dominant predictor.
- At national levels, indirect exposure variables (news coverage frequency, social media use) become dominant predictors, superseding direct experience.
Sociodemographics: Lower income and lower education levels are strong predictors of overestimation, but these do not attenuate the effects of exposure variables.

4. Key Contributions

Multi-Scale Analysis: First study to systematically map overestimation across four nested geographical scales, revealing a distinct transition from experience-based inference (local) to media-based inference (national).
Mechanism Disentanglement: Differentiates between the roles of direct social contact and various forms of indirect media exposure (traditional news vs. social media, tone vs. frequency).
PoC Perspective: Provides empirical evidence that minority groups also overestimate their own size, challenging the view that this is solely a majority-group bias.
Methodological Rigor: Utilizes a purpose-built survey with ground-truth census matching and advanced machine learning (Random Forest + SHAP) to model non-linear interactions between variables.

5. Significance and Implications

The "Illusion of Diversity": The findings suggest that overestimation creates a false perception that diversity goals have been achieved. This "illusion" can reduce public support for policies designed to promote equity and representation.
Media Literacy: The divergence between traditional news (corrective) and social media (reinforcing) highlights the need for critical media literacy. Social media's combination of news diffusion and interpersonal contact creates a perceived environment more diverse than reality.
Policy Relevance: Understanding that national perceptions are driven by media narratives rather than local reality is crucial for designing communication strategies that correct demographic misperceptions without triggering defensive reactions.
Future Research: The authors note limitations (cross-sectional design, aggregation of PoC groups) and suggest future work using longitudinal designs and cognitive modeling to understand the underlying judgment processes.