The guide says identities are intersectional and may include age, nationality, religion, culture, gender, sexuality, race, ethnicity and social and economic class. The key idea is that the combination is a position of its own, not the sum of its parts.
One system passing two fairness tests taken separately and failing badly at the intersection.
Interactive diagram
Explore the labelled diagram, charts and maps for this topic in full study mode.
Free preview
This is the free notes preview
You're reading the free notes. Aimnova Pro unlocks the full study experience — and you can try it with your first topic free to keep:
- FlashcardsLock in vocabulary and key terms with spaced repetition.
- Practice questionsAnswer exam-style questions and get instant AI marking.
- Mock exams & past-paper vaultSit full mocks and see exactly how examiners award marks.
- Personalised study planA daily plan built around your exam date and weak areas.
This is not a purely theoretical idea. It changes what a fair test looks like, and published audits of face-analysis systems found exactly the pattern the concept predicts.
What follows in practice
- Testing one characteristic at a time can pass while the system fails for people at the intersection.
- Small groups disappear in an average. A 1% error rate overall can be a 30% error rate for a group that is 2% of the data.
- Training data is thinnest exactly there. The fewer examples of a combination, the worse the model handles it.
- The people affected are least likely to be consulted, which is why the failure reaches deployment.
See how examiners mark answers
Access past paper questions with model answers. Learn exactly what earns marks and what doesn't.
Real-world examples you can name
COMPAS recidivism scoring — ProPublica investigation May 2016
A commercial risk-scoring tool used in some US courts to estimate how likely a defendant was to reoffend. Journalists found that among people who did not go on to reoffend, Black defendants were roughly twice as likely as white defendants to have been labelled high risk. The company disputed the measure of fairness used, and the argument that followed showed that competing definitions of a fair algorithm cannot all be satisfied at once.
Who it affected: Defendants whose bail and sentencing decisions were informed by the score.
Amazon's scrapped CV-screening tool — reported October 2018
An experimental hiring tool was trained on a decade of CVs submitted to the company, most of them from men. It learned to downgrade CVs containing the word 'women's' and graduates of two women's colleges. The project was abandoned.
Who it affected: Women applying for technical roles, and every organization that assumed historical data was neutral.
Aadhaar, India's biometric ID system — launched 2009; Supreme Court ruling September 2018
A national identity number linked to fingerprints and iris scans, used to authenticate access to services. The Supreme Court upheld the scheme for welfare and tax purposes but struck down its use by private companies, and reporting has linked authentication failures to people being denied rations.
Who it affected: Over a billion enrolled residents, and in particular those whose fingerprints scan poorly — manual labourers and older people.
The thread: In all three the harm is concentrated on a group defined by more than one characteristic, and in all three the system was assessed against one characteristic at a time.
How this is tested — this concept gives you a specific, technical counter-argument that most answers do not reach for. It comes up two ways:
Paper 1 — structured question
- Part c: evaluate a system's fairness
- Part b: explain why a fairness test may miss a problem
Paper 2 — source-based question
- Q4: synthesise sources on who a system disadvantages
- HL Paper 3: recommending an intervention that must be tested
The trap: listing characteristics: The point is not that identity has many parts. It is that the combination behaves differently from any of the parts alone.
Explain one reason why a system can pass a fairness test and still disadvantage a group.
Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.
Suggest two ways a developer could test an automated system for intersectional bias.
Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.