Skip to content

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

Kestrel Lending, a bank, uses a machine learning model to approve personal loans. An internal review finds that applicants from certain postcodes are declined far more often, even though postcode is not an input. The model uses a proxy variable highly correlated with postcode. Which ethical and governance issue does this most directly illustrate?

This illustrates algorithmic bias through proxy variables. Even though postcode is excluded, a correlated variable reproduces its effect, leading to unfair outcomes for certain groups. Governance should require bias testing, fairness monitoring and review of input variables, not just removal of the sensitive field.

  1. AAlgorithmic bias arising from proxy variables in training dataCorrect
  2. BModel drift caused by changes in interest rates
  3. CLack of processing capacity in the bank's cloud infrastructure
  4. DOverfitting caused by too few input variables

Explanation

Removing a protected or sensitive attribute does not remove bias if other variables act as proxies for it. The outcome is discriminatory and creates fairness, reputational and legal risk. Model drift concerns changing data patterns over time, not unfair outcomes linked to a proxy.

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