itsual Data Ethics
Overview
Data ethics asks what should be done with data and models, not only what can be done. It covers privacy, fairness, transparency, and potential harm.
When to Use
- Projects with personal or sensitive data
- Models affecting access to opportunities (credit, jobs, services)
- Publishing or sharing analytical results externally
- Designing data collection practices
Core Practices
- Minimize data collected and retained for the purpose
- Assess fairness and disparate impact where decisions affect people
- Obtain and respect appropriate consent and lawful basis
- Avoid re-identification risks in “anonymized” releases
- Communicate uncertainty and limits honestly
- Escalate dual-use or harm concerns early
Principles
- Legal compliance is necessary but not sufficient for ethics
- Aggregate results can still harm groups
- “The data said so” does not remove human responsibility
- Document decisions about sensitive trade-offs
Verification
- Purpose limitation and minimization are considered
- People-impacting uses have fairness/privacy review
- Communications do not overclaim certainty