Définition
Accountability in AI means not blaming a controversial decision on an “algorithm” without explanation. The system must have an owner, control rules, logging, a procedure for correcting errors, and a clear way to challenge the result. This is especially important in finance, medicine, hiring, education and government services.
Exemple
If an automated system rejects a loan application, the company must understand why this happened, who controls the model, and how a person can challenge the decision.
Pourquoi c'est important
The term helps distinguish mature AI solutions from risky ones: a useful service not only produces results, but also shows who is responsible for quality and consequences.
Fonctionnement
Responsibility is built through roles, internal policies, auditing, model documentation, monitoring, risk assessment and the possibility of human review.
Où c'est utilisé
- AI risk management
- algorithm audit
- policies for the use of AI in the company
Limites
Documents by themselves do not protect against errors. Accountability only works when people have the power to stop, fix, or review the system.
FAQ
Why is “Accountability” useful to know?
The term helps distinguish mature AI solutions from risky ones: a useful service not only produces results, but also shows who is responsible for quality and consequences.