Key Idea: An algorithm is a finite, ordered set of unambiguous instructions. The examinable point is that sorting, searching, filtering and classifying do very different things to a person, and only one of them is visible afterwards.
Paper 1
- Part a: define an algorithm, or identify its components
- Part b: explain why a threshold is a decision somebody made
- Part c: evaluate algorithmic decision-making about people
Paper 2
- Q2: analyse a source's claim about an automated decision
- Q3: compare sources on consistency and discretion
- Q4: synthesise sources on hiring or scoring
Everywhere else
- AI: a learned rule is still an algorithm nobody wrote
- Power: the threshold is where power becomes a number
- Health and politics: filtering decides who is reached
Four operations, four different consequences
| Operation | What it does to a person | Visible afterwards? |
|---|---|---|
| Sorting | Changes their position; everybody stays | Yes — they are still in the list |
| Searching | Answers a question about one of them | Yes |
| Filtering | Removes them from what follows | No — the excluded and the absent look identical |
| Classifying | Attaches a label that travels | Partly — the label outlives the estimate |
Filtering leaves no trace for anybody downstream, so the effect of moving a threshold cannot be observed from the outcomes. The only way to examine it is to inspect the rule before it runs.
Consistency is a gain and a risk
Both halves are true at once
- The gain — the outcome no longer depends on who was on duty, which removes a real form of unfairness.
- The risk — a rule that is wrong for a particular person is wrong for them every single time, with nobody able to make an exception.
- The consequence — errors stop being random and become systematic, which makes them larger in total and harder to notice one at a time.
Important: The outcome is produced by a chain — the data, the objective, the deployment, the use — and no single link decided the specific result, so each participant can truthfully say they did not choose it.
Exam-style questions
Compare the effect of sorting and classifying on the people a system processes.
🔒 Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.
Justify the claim that an algorithm should be examined before it runs rather than judged by its outcomes.
🔒 Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.
Quick check
Which operation is invisible? Filtering. Nobody downstream can tell the excluded from the absent.
Why is a threshold a decision? Somebody chose the number, and it appears in the system as a fact rather than as a judgement.
Why is consistency also a risk? An error stops being random and becomes systematic — wrong for the same person every time.
Where does responsibility go? Into a chain where no single link chose the specific outcome, so the harm has no owner.
Exam tips
- Say which operation a system performs. 'The algorithm decides' is not an answer.
- Filtering is the one with teeth. Reach for it whenever a question is about who was excluded.
- Publishing code is not transparency if the people affected cannot read it.