Data Ethics

Data Is Never Neutral

Every dataset reflects choices: what to measure, what to exclude, how to categorize, and who gets to decide. These choices are not technical — they are ethical. The US Census's racial categories have changed significantly over time, reflecting evolving social constructions of race. Data about 'crime' measures policing patterns as much as criminal behavior. Data about 'achievement' in schools reflects socioeconomic conditions as much as teaching quality.

Catherine D'Ignazio and Lauren Klein's 'Data Feminism' argues that data science must interrogate power: who collects the data, who benefits from it, who is harmed by it, and whose perspectives are missing. A dataset about healthcare outcomes that excludes undocumented immigrants renders their health needs invisible to policymakers.

Consent, Privacy, and Power Asymmetry

Most people have no meaningful ability to consent to data collection. Terms of service agreements are unreadable by design — Facebook's terms are longer than the US Constitution. Clicking 'I agree' is not informed consent in any meaningful sense.

The power asymmetry is stark: companies and governments know enormously more about citizens than citizens know about them. This asymmetry enables manipulation — targeted advertising that exploits vulnerabilities, political micro-targeting that exploits fears, and predictive policing that reinforces existing biases.

Data ethics requires asking not just 'can we collect this data?' but 'should we?' Not just 'is this analysis technically correct?' but 'is it fair?' Not just 'does this algorithm work?' but 'who does it work for and who does it harm?'