💻 computer science

Causal evidence of racial and institutional biases in accessing paywalled articles and scientific data

Through a combination of large-scale analysis and randomized audit experiments, this study demonstrates that racial and institutional biases in informal channels significantly hinder researchers from the Global South from accessing paywalled articles and scientific data, thereby perpetuating structural inequities in scientific progress.

Original authors: Hazem Ibrahim, Fengyuan Liu, Khalid Mengal, Wisam Alshaibi, Aaron R. Kaufman, Yasir Zaki, Talal Rahwan

Published 2026-07-09

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Hazem Ibrahim, Fengyuan Liu, Khalid Mengal, Wisam Alshaibi, Aaron R. Kaufman, Yasir Zaki, Talal Rahwan

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 the world of scientific research as a massive, bustling library. In this library, the most important books (scientific papers) and the raw materials used to write them (data sets) are the keys to building new knowledge. Ideally, this library should be open to everyone, like a public park where anyone can walk in and learn.

However, this paper argues that in reality, the library has invisible bouncers, expensive membership fees, and a "back door" that only some people can use. The researchers behind this study wanted to find out: Does it matter who you are or where you come from when you try to get through that back door?

Here is a breakdown of their findings using simple analogies:

1. The Two Libraries: "Rich" vs. "Poor" Neighborhoods

The researchers first looked at the "official" way to access science.

The Global North (Rich Neighborhoods): Universities here have expensive subscriptions. It's like having a VIP pass that lets you walk right through the front door into the library.
The Global South (Poor Neighborhoods): Universities here often can't afford the VIP pass. Their researchers are locked out of the front door.

The Finding: When researchers in the "poor" neighborhoods try to get in, they are far less likely to have a key. They hit "paywalls" (locked doors) much more often than their wealthy counterparts.

2. The "Back Door" Strategy: Asking the Author

Since they can't use the front door, researchers in the Global South often try the "back door." They send an email directly to the scientist who wrote the paper, saying, "Hi, I'm a student. Can you please send me a copy of your paper or the data you used?"

The researchers wondered: Does it matter who sends the email?

To test this, they played a game of "fake student." They created fictional PhD students with different names and university backgrounds and sent thousands of emails asking for papers and data.

The Experiment: They sent emails from a student named "Karl Wagner" (German/White) and a student named "Oluwasen Adeyemi" (Nigerian/Black), among others. They also varied the university the student claimed to attend (a top-tier elite school vs. a lower-ranked one).

The Results:

The Name Game: The emails from the German/White student got a reply about 15% of the time. The emails from the Nigerian/Black student only got a reply about 11% of the time. The Pakistani/South Asian student did even worse.
The School Game: When asking for data (the raw numbers), it mattered where the student claimed to go to school. Emails from elite universities got more replies than those from lower-ranked ones.
The "No" Rate: Even for the most privileged sender, the success rate was low (only about 1 in 7 emails got a reply). For the marginalized senders, it was even worse (about 1 in 9).

The Metaphor: Imagine you are asking a stranger for a ride. If you look like them or come from their neighborhood, they might roll down the window and say "Sure." If you look different or come from a place they don't know, they might not even look up from their phone. The study found that scientists are doing exactly this: they are less likely to even reply to the email if the sender has a name or background associated with the Global South.

3. The "Data" vs. "Paper" Difference

The study found a subtle difference in how people are treated depending on what they ask for:

Asking for a Paper: This is like asking for a recipe. It's low effort. Here, race was the biggest factor. If your name sounded "foreign," you were less likely to get a response.
Asking for Data: This is like asking for the raw ingredients and the chef's secret notes. It's higher effort and riskier for the author. Here, institutional prestige mattered more. Authors seemed to use the university name as a "trust badge." If the student claimed to be from a top school, they were more likely to get the data, regardless of their race.

4. The Ripple Effect: What Happens When You Can't Get In?

The researchers didn't just stop at the emails; they looked at the long-term damage.

The Narrow View: Because researchers in the Global South can't access the "locked" papers, they end up citing (referencing) fewer of them. It's like trying to paint a picture but only having access to a few colors. Their work ends up being based on a narrower range of knowledge.
The Impact Gap: Papers that rely on a wider range of knowledge tend to be more influential. By being locked out, Global South researchers are inadvertently making their own work less visible and less impactful, not because their ideas are bad, but because they couldn't see the full picture.

5. The "Conditional" Gatekeepers

In the interviews, some researchers described a frustrating reality. When they did get a reply, it wasn't always a simple "Here is the file."

Sometimes the author said, "I'll give you the data if you make me a co-author on your paper."
Sometimes they said, "I'll give you the data if you pay me $1,000."
Sometimes they just ignored the email.

The Metaphor: It's like a bouncer at a club who says, "You can't come in the front door. You can try the back door, but only if you know the right person, look the right way, or are willing to pay a bribe."

The Bottom Line

The paper concludes that the scientific world is not as open as it claims to be. Even though scientists say they want to share knowledge, informal barriers (like who gets a reply to an email) are creating a system where:

Race matters: Being Black or South Asian makes it harder to get a response.
Geography matters: Being from the Global South makes it harder to get a response.
Prestige matters: Being from a top university helps, but it doesn't fix the racial bias.

The authors argue that we can't just rely on scientists being "nice" and sharing their work via email. We need structural changes—like making all data and papers automatically available to everyone—so that no one has to knock on a door that might be locked against them.

Technical Summary: Causal Evidence of Racial and Institutional Biases in Accessing Paywalled Articles and Scientific Data

Problem Statement
Scientific progress relies on the cumulative ability of researchers to access, scrutinize, and build upon prior work. However, a significant portion of scientific literature remains behind paywalls, and underlying datasets are often restricted to "reasonable requests" to authors. While the open science movement aims to remove these barriers, informal channels (e.g., emailing authors directly) have emerged as a primary workaround for researchers lacking institutional subscriptions. The central problem addressed by this study is whether these informal channels are subject to racial and institutional biases, potentially perpetuating structural inequities between researchers in the Global North and the Global South. The authors investigate if access to knowledge is mediated not just by the availability of resources, but by the identity and affiliation of the requester.

Methodology
The study employs a mixed-methods approach combining survey data, qualitative interviews, large-scale observational bibliometric analysis, and two randomized audit experiments.

Study 1 (Survey and Interviews):
- Survey: A global survey of 376 researchers (216 from the Global North, 160 from the Global South) across 76 countries assessed institutional access to publisher databases, frequency of paywall encounters, coping strategies, and self-reported success rates when emailing authors.
- Interviews: 13 semi-structured interviews with researchers from the Global South provided qualitative insights into the lived experience of navigating access barriers, including the use of informal networks and the geopolitical frictions encountered.
Study 2 (Observational Citation Analysis):
- Utilizing the OpenAlex database (250+ million articles), the authors analyzed citation patterns to determine if researchers from the Global South cite paywalled papers at different rates than those from the Global North.
- A neural embedding model (Doc2Vec) was used to measure the "knowledge breadth" of reference lists, assessing whether limited access to paywalled articles correlates with narrower topical focus and reduced scholarly impact.
Study 3 (Audit Experiment: Paywalled Articles):
- Design: A randomized controlled trial (RCT) sending 18,000 emails to corresponding authors of paywalled papers.
- Variables: The sender's identity was manipulated in a $3 \times 3 \times 2$ $3 \times 3 \times 2$ factorial design:
  - Race/Nationality: German/White, Nigerian/Black, Pakistani/South Asian.
  - Institution: Top-50, 501–600, and 1000+ ranked universities.
  - Purpose: Journal article vs. class project.
- Outcome: The primary metric was the rate of substantive replies.
Study 4 (Observational Data Citation Analysis):
- Using the Semantic Scholar database, the authors retrieved Data Availability Statements (DAS) for 200,000 papers and classified 4 million citation contexts using a custom deep learning classifier to identify citations made specifically for data reuse.
- The study examined whether papers with data available "upon request" received fewer data citations, and if this effect was amplified when the citing paper originated from the Global South.
Study 5 (Audit Experiment: Dataset Requests):
- Design: A second RCT sending 16,000 emails to authors of papers with "upon request" data statements.
- Variables: A $2 \times 2 \times 2 \times 2$ $2 \times 2 \times 2 \times 2$ factorial design:
  - Race/Nationality: English/White vs. South African/Black.
  - Institution Country: England vs. South Africa.
  - Institution Rank: High (150–155) vs. Low (1001–1200).
  - Purpose: Journal article vs. class project.
- Outcome: Reply rates and the nature of the response (positive, conditional, or negative).

Key Results

Access Disparities (Study 1): Global South researchers reported significantly lower institutional access to major publisher databases (e.g., 62% access to Elsevier vs. 83% in the Global North) and encountered paywalls more frequently. They relied more heavily on informal channels but reported significantly lower success rates when emailing authors for papers (26% success vs. 44% in the Global North) and datasets (15% vs. 34%).
Citation Patterns (Study 2): Researchers in the Global South cited paywalled articles at significantly lower rates (61% of references vs. 66% in the Global North). Papers with higher proportions of paywalled references were associated with broader knowledge breadth, suggesting that restricted access limits the scope of literature reviews.
Causal Evidence of Racial Bias (Study 3): In the article request audit, racial identity significantly influenced response rates. Relative to the German/White sender, Nigerian/Black senders received 31.6% fewer replies, and Pakistani/South Asian senders received 37.4% fewer replies. Institutional prestige did not significantly affect article request responses, suggesting the bias was primarily racial/taste-based rather than prestige-based for articles.
Data Citation Disparities (Study 4): Papers with data available "upon request" were significantly less likely to be cited for data reuse. This penalty was disproportionately higher for citations from the Global South, indicating a compounding effect of geographic and access barriers.
Causal Evidence of Racial and Institutional Bias (Study 5): In the dataset request audit, both racial identity and institutional affiliation independently predicted response rates.
- South African/Black senders received 22.3% fewer replies than English/White senders.
- Requests from South African institutions received 20.2% fewer replies than those from English institutions.
- Requests from lower-ranked institutions received 15.7% fewer replies than those from higher-ranked institutions.
- Unlike Study 3, institutional cues were a salient factor for data requests, suggesting authors rely on institutional prestige as a heuristic for trustworthiness when sharing data.

Significance and Claims
The paper claims to provide the first large-scale experimental evidence that informal gatekeeping in science is shaped by both racial and institutional biases. The authors argue that while open science policies aim to democratize access, the reliance on informal, discretionary channels (emailing authors) creates a new site of inequality.

Key claims include:

Structural Inequity: Informal access channels are not neutral; they reproduce existing hierarchies of race and geography. Researchers from the Global South and marginalized groups face systematic barriers even when formal access is unavailable.
Epistemic Vulnerability: The findings suggest that publication does not guarantee equal accessibility. Knowledge remains under the informal control of authors, leading to downstream effects where Global South researchers have narrower knowledge bases and reduced scholarly impact.
Policy Implications: The authors argue that ensuring equitable access requires more than just fair treatment in informal requests. They call for stronger, enforceable data-sharing mandates, expanded open access policies, and institutional support to reduce dependence on informal requesting altogether. The current system, where data sharing is often transactional or conditional, shifts the scientific enterprise away from Mertonian ideals of communality and disinterestedness.

The study concludes that addressing these disparities is essential for a truly inclusive scientific ecosystem, as current informal mechanisms disproportionately exclude researchers who are often closest to global challenges such as climate adaptation and public health.