Four rules for AI in health care
A clinic uses an AI to choose who gets reminder calls. How can we tell it is fair?
- Respect for autonomy: people understand how the AI decides. They have a say.
- Do not harm: avoid hurting anyone. If every choice risks harm, pick the least.
- Maximum benefit: do more than avoid harm. Aim to do the most good.
- Justice: share the good and the bad fairly, whatever a person's background.
The rule
We check fairness by asking four questions, one for each rule.
Bioethics is a way to judge choices in health care. It is a sector-based framework: a plan for one field, health care.
Bioethics has four rules: autonomy, do not harm, maximum benefit and justice.
These four are the best known. Other ways to judge exist.
- Autonomy — a person's right to understand a decision and have a say.
- Do not harm is also called non-maleficence.
- Maximum benefit is also called beneficence.
- Justice — every person gets a fair share, whatever the background.
One made-up case
A clinic AI decides which children get a reminder call. The call is for a free check-up. The AI learned from who came on time last year. Children from far villages came late, because buses are rare. So the AI calls them less.
| Rule | The question | One fix |
|---|---|---|
| Autonomy | Can parents see how it picks? | Tell parents how the AI picks who gets a call. |
| Do not harm | Could a child miss a check-up? | Check who the AI skips. Fix it before use. |
| Maximum benefit | Can it help more children? | Learn from who needs a check-up, not who came on time. |
| Justice | Is it fair to every village? | Give every village the same chance of a call. |
Watch out!
These rules help us judge an AI. They do not tell you what to do about your own health. Ask a doctor or a trusted grown-up.
Flip the three switches below. Watch which rule each fix follows. Do not harm needs two fixes. The AI must learn from who needs a check-up. It must also give every village the same chance.
Loading lab…