ACCA Strategic Professional · Strategic Business Leader · Machine learning, AI and robotics
Tamsin Insurance deploys AI to approve motor claims automatically. Auditors find the model is highly accurate overall but cannot explain individual rejections, and rejection rates are noticeably higher for one postcode area. Which response best reflects sound governance of AI?
The sound response is to add human oversight and explainability requirements and to test the model for bias. High overall accuracy does not justify unexplained or potentially discriminatory rejections, while abandoning AI entirely or making it more complex would be disproportionate or worsen opacity.
- AIntroduce human oversight and explainability requirements, and test the model for bias before continuing to rely on itCorrect
- BContinue unchanged because overall accuracy is high
- CWithdraw all AI use permanently because algorithms cannot be governed
- DMake the model more complex so that it learns to remove the postcode effect itself
Explanation
Overall accuracy does not remove the risk of unfair outcomes or the need to justify decisions to customers and regulators. Oversight, explainability and bias testing address both issues. Permanent withdrawal is disproportionate, and added complexity worsens opacity.
Did you get it right without looking?
One question tells you little. A timed set on Machine learning, AI and robotics shows your real accuracy, how long you take and where you lose marks.
More Machine learning, AI and robotics questions
- Dalmore Health uses a deep learning system to prioritise patients for specialist referral. Clinicians cannot explain why the system ranks pa…
- Tamsin Foods deploys software robots that log into its systems, copy supplier invoice data into the ledger and follow fixed rules, with no a…
- Halden Retail uses a machine learning model to shortlist job applicants. The model was trained on ten years of past hiring decisions, during…
- Dalton Insurance feeds its entire customer database, with no outcome labels, into an algorithm that groups customers into segments with simi…
- Orlan Logistics plans to replace 40% of its warehouse workforce with robots and AI scheduling. The CEO argues the decision is purely commerc…
- Calder Bank wants to reduce card fraud losses. Fraud patterns change constantly and are not fully captured by any fixed list of rules. The b…