Passage: 'Algorithmic credit scoring models often replicate historical inequities when trained on biased legacy datasets. Even when sensitive demographic variables are expunged, proxy attributes such as postal code or educational pedigree can indirectly recreate discriminatory outcomes.'

How can algorithmic bias persist even without demographic inputs?

Answer and explanation

Correct answer: C. Correlated non-demographic variables act as proxies for protected characteristics

  1. Step 1: Examine how machine learning models sustain biased patterns when direct demographic variables are removed.
  2. Step 2: The passage notes that "proxy attributes such as postal code or educational pedigree can indirectly recreate discriminatory outcomes," because these alternative variables correlate strongly with protected demographic groups.
Answer
Correlated non-demographic variables act as proxies for protected characteristics.

Practise more Multiple Choice: Academic Reading Comprehension.

Free questions with an explanation for every one, and full timed PTE Academic mock tests.

Passage: 'Algorithmic credit scoring models… | PTE Academic