Should AI and data-science students learn about bias and fairness?
Yes. Students of AI and data science should learn how bias can arise during data collection and labeling, modeling, and deployment, as well as how fairness can be understood and assessed. Because definitions of fairness depend on the purpose and social context, and different measures may conflict, students should consider not only technical tests but also ethical and social impacts, limitations, and how results are communicated.
Key points
- Examine possible sources of bias at the data, model, and deployment stages.
- Fairness has multiple definitions and measures, so judgments must reflect the purpose and context.
- Consider affected people, uncertainty, and technical limitations as well as predictive performance.
Things to check
- The content and depth vary by university, faculty, course, and academic year, so specific curricula should be checked in official information.