Skip to content

ACCA Strategic Professional · Strategic Business Leader · Machine learning, AI and robotics

Brightpath Recruitment uses an AI tool trained on ten years of its own hiring decisions to shortlist candidates. An internal review finds the tool systematically ranks women lower for senior roles, because past managers hired mostly men. What is the most accurate diagnosis and implication?

The tool has learned historic bias from the training data and so reproduces discrimination. The implication is that the board should require bias testing, better data and human oversight. Greater computing power or changing hosting would not remove bias already embedded in the model.

  1. AThe training data embeds historic bias, so the model reproduces discrimination and the board should require bias testing and human oversightCorrect
  2. BThe model is malfunctioning technically, so it should simply be given more processing power
  3. CThe result is acceptable because the model is objective and has no human involvement
  4. DThe problem arises only because the tool is cloud-hosted, so moving it in-house would solve it

Explanation

Models learn patterns in historic data, including biased human decisions, so the output reflects past discrimination. The remedy is governance: bias audits, representative data, and human review of outcomes. More power or relocating hosting does not change what the model has learned, and 'objective' is wrong because bias is inherited.

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.

Practise Machine learning, AI and roboticsFree diagnostic mock

More Machine learning, AI and robotics questions