Skip to content

ACCA Strategic Professional · Advanced Performance Management · Data science and analytics

Calder Insurance uses a neural network to price motor policies. The model is accurate, but regulators ask the finance director to explain why individual customers receive particular premiums. The network cannot show the reasoning behind outputs. Which issue does this best illustrate, and what is the most appropriate response?

This is the black-box problem: the model's reasoning cannot be explained, creating regulatory, fairness and accountability risk. The appropriate response is to use explainable AI techniques or a more interpretable model, with documented governance and human oversight of decisions.

  1. AData latency; move to real-time data feeds
  2. BThe black-box problem; use explainable AI techniques or a more interpretable model and keep human oversightCorrect
  3. COverfitting; add more training variables
  4. DData redundancy; delete duplicate customer records

Explanation

Inability to explain how outputs were reached is the black-box or transparency problem, a key governance and ethical concern for AI. Responses include explainability tools, simpler interpretable models, documentation and human review. The other options address unrelated technical issues.

Did you get it right without looking?

One question tells you little. A timed set on Data science and analytics shows your real accuracy, how long you take and where you lose marks.

Practise Data science and analyticsFree diagnostic mock

More Data science and analytics questions