How does data reflect power
The Data Is Never Just the Data
A survey result can look wonderfully clean on a slide: 62 percent chose option A, the sample is neatly tabulated, and someone has added a reassuring blue circle around the number. But before treating that result as a description of reality, I want to know who was able to answer, which categories were available, and what the question quietly assumed about people’s lives.
That is the starting point I took from Catherine D’Ignazio and Lauren F. Klein’s Data Feminism, an open-access book about power, data, and the systems built around them. The authors make a useful case against the phrase “the data speaks for itself.” Data does not speak. People decide what to collect, how to classify it, which gaps are acceptable, and whose uncertainty gets smoothed away.
A spreadsheet has a point of view In product research, the most consequential decisions often happen before anyone opens the analysis tool. A team decides that “household” means one address. A form offers only two gender options. A public service measures successful applications but not the number of people who gave up halfway through. None of these choices are neutral technical details. They shape what becomes visible.
This does not mean every dataset is useless or that researchers should abandon measurement. It means the method needs an account of its own limits. A percentage without context can create false confidence, particularly when it is used to justify an automated decision or to redesign a service for people who were barely represented in the original research.
Power is also about who gets to refuse One point I found especially practical is the book’s attention to participation and resistance. Ethical data work is not only about inviting people into a study. It is also about noticing who cannot safely participate, who has no meaningful way to opt out, and who may be harmed when a sensitive detail is converted into a permanent record.
That matters in civic technology, where the product may be a benefits portal, a housing system, or a tool used by a city employee. “More data” can sound like an obvious improvement until you ask whether residents understand the collection, whether the categories fit their situation, and whether the resulting system will make a confusing service easier or merely make the confusion more efficiently documented.
A better research question The most useful takeaway is not a checklist. It is a habit of asking where power enters the process. Who defined the problem? Who benefits from this measurement? Who is missing? What would change if the people most affected could challenge the categories instead of merely selecting from them?
Those questions slow down the rush toward an AI feature, which is probably healthy. They also make research more honest. A model can be technically impressive and still be built on a narrow account of people’s lives. A dashboard can be accurate and still point decision-makers toward the wrong priority.
I am linking to the book because it is unusually accessible for a serious discussion of data ethics, and because it gives language to a feeling many researchers encounter: the number is not lying, exactly. It is just not telling the whole story.