Simpson's Paradox in University Admissions: How Department Choices Shape Perceived Bias

Overview

In a thought experiment about a future cat-topia university, the video shows how simple department level statistics can mask bias at the university level via Simpson's paradox. It uses astronomy and physics admissions, and a Berkeley historical case, to illustrate how the distribution of applicants across departments can create misleading overall results.

• Simpson's paradox explains why a group can appear favored overall even when there is no bias within each department.
• The key idea is the allocation of applicants to departments drives university level acceptance rates.
• The Berkeley example reveals societal factors that channel applicants into less funded or more competitive departments, producing unequal representation despite fair departmental processes.

The video ends with a call to recognize and address biases beyond admissions, highlighting the broader social context that shapes opportunities in science and education.

Introduction

The video presents a thought experiment in which a hypothetical university with two departments, astronomy and physics, treats cats and humans as applicants. Across departments, cats and humans experience different acceptance counts, which, when aggregated, yields overall acceptance rates that seem biased against one group. This setup introduces Simpson's paradox, a statistical phenomenon where trends that appear in several different groups disappear or reverse when the groups are combined.

Simpson's Paradox in Admissions

The core idea is straightforward: if more cats apply to the astronomy department and that department has a higher overall acceptance rate, while more humans apply to physics and that department has a lower acceptance rate, the university as a whole can show unequal acceptance by species even if each department treats applicants fairly. The cats and humans are each judged within their own department, but the distribution of applicants across departments skews the university-wide numbers. This illustrates Simpson's paradox in a concrete, accessible way.

A Real World Parallel: Berkeley in the 1970s

The video connects the thought experiment to a real historical example from Berkeley in the 1970s. At the time, the graduate school admitted 44% of men and 35% of women overall, which suggested a gender bias. Careful departmental analysis, however, showed no bias within departments in the admissions review process. Women tended to apply to departments with less funding and fewer places, such as English, while men tended to apply to more competitive departments like engineering. The paradox here is not in the departmental reviews but in the uneven distribution of applicants across departments, which produced an overall gender imbalance at the university level.

Beyond Admissions: Societal Bias and Policy Implications

The video argues that Simpson's paradox reveals how institutional structures can propagate bias even when internal processes are fair. It invites readers to consider how advertising smaller, more competitive departments to certain groups, or perceptions about departmental cultures, might influence who applies where. The Berkeley study is cited as a cautionary tale about larger societal forces shaping academic fields, funding, and opportunity for different demographic groups.

Takeaway and Call to Action

Ultimately, the paradox emphasizes that a lack of obvious bias within a department does not guarantee equality across a university or system. Real-world bias can be driven by where people are steered to apply, what fields are perceived as welcoming, and what funding and opportunities are available. The video calls for ongoing awareness and proactive efforts to address biases across the entire educational ecosystem, not just at the point of admission.