Power, Data, and Studying Up

For each of the hypothetical scenarios below, please:

  1. Briefly analyze the scenario in terms of the four domains of the matrix of domination. Which of these domains are operating in the scenario provided? There are probably multiple, but not necessarily all four.
  2. Suggest a way in which data scientists could “study up” (Barabas et al. 2020) by collecting and analyzing data about the powerful or privileged actors in the scenario.

A concise paragraph for each scenario is plenty.

Scenarios

  1. Landlords use an algorithm to predict “payment reliability” of prospective tenants in apartment complexes. The algorithm gives lower reliability scores to prospective tenants moving from predominantly Latinx/Hispanic neighborhoods than it does to prospective tenants moving from predominantly white neighborhoods.
  2. Police use predictive policing algorithms that encourage them to allocate officers to neighborhoods that have historically had high arrest rates. A local politician tweets that these neighborhoods are “blighted.”
  3. Medical professionals use a health score to determine which patients should receive palliative care. The score includes an assessment of pain levels as one of its inputs. Doctors are known to systematically underestimate the pain levels of Black patients in comparison to white ones.



© Phil Chodrow, 2025

References

Barabas, Chelsea, Colin Doyle, Jb Rubinovitz, and Karthik Dinakar. 2020. “Studying up: Reorienting the Study of Algorithmic Fairness Around Issues of Power.” In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, 167–76. Barcelona Spain: ACM. https://doi.org/10.1145/3351095.3372859.