AI Ethics
The field asking not whether AI can do something but whether it should — and who bears the consequences when it does.
Reviewed July 15, 2026Stable
When not to use it
- (Not applicable — AI ethics is a lens applied throughout, not an optional module. The failure mode is treating it as a box to tick at the end.)
Reach for something else instead
- Regulatory compliance as the enforceable floor — necessary but not sufficient for ethics.
- Value-sensitive design as a methodology for building values in from the start.
Read more on the blog
- Nine cases, and none was fixed by a better modelTerritory 6 closes. Nine documented cases, three containing no AI at all, and not one where the remedy that worked was a more accurate system.
- Robodebt: losing quietly to avoid losing publiclyA Royal Commission found the scheme unlawful, crude and cruel. The tribunal had been ruling against it for years, and the department never appealed, so no precedent was ever set.
- Horizon: the law presumed the computer was rightHundreds prosecuted on the output of an accounting system later found not to be robust. No machine learning was involved, which is precisely why it belongs in this record.
- The Dutch benefits scandal: the rule, not the modelAround 26,000 families were wrongly accused of fraud and a government resigned. The parliamentary inquiry did not blame the algorithm. What it found is more useful and less quoted.
Further reading
- Kleinberg, Mullainathan & Raghavan (2016), Inherent Trade-Offs in the Fair Determination of Risk Scores — the impossibility result for fairness definitions.
- Mitchell et al. (2019), Model Cards for Model Reporting — a practical accountability tool.
- Jobin, Ienca & Vayena (2019), The global landscape of AI ethics guidelines — a survey of the principles that recur worldwide.
Primary sources, listed so you can check the claims on this page rather than take them on trust.
Where people go wrong
- Treating ethics as a final-stage checklist rather than a throughout-the-lifecycle question.
- Assuming "fair" is well-defined — several fairness metrics are provably incompatible.
- Mistaking algorithmic objectivity for neutrality — the choices in data and design carry values.
At a glance
FieldSafety & Ethics
Core questionnot can but should
Pillarsfairness, transparency, accountability, privacy, safety
Hard truthvalues genuinely conflict
DifficultyIntermediate
Where this sits
1 concept come first. Understanding it opens up 20 more.
1Levelsteps in
1Needs firstconcepts
20Opens up6% of 310
3Areascrosses fields
Learn these firstArtificial Intelligence
Computed from the prerequisite graph, not assigned. How this works