A public-beach locker kiosk prompt asks an LLM for guest names for test data. The reply returns only English-language given names even though the city serves several language communities. Which defect does this illustrate?
Select an answer to reveal the explanation.
Short Explanation
If the training diet skewed English, the name menu will too. That skewed output is bias from training data, including language under-representation.
Full Explanation
When a public-beach locker kiosk prompt asks an LLM for guest names for test data and the reply returns only English-language given names even though the city serves several language communities, the defect illustrated is bias originating in the data the model was trained on. Civic beach services must test representation across the communities they serve, not a skewed name diet. A hallucination inventing a locker fee not on the rate board fails as the wrong class when fees were not invented—names were skewed. A reasoning error reversing checkout before check-in fails because the stem is about demographic skew, not step order. Non-determinism from sampling alone fails as the primary label for systematic under-representation. Exam caveat: training-data bias shows up as skewed synthetic populations. Operational check: review name lists against the city’s language communities, regenerate with explicit diversity constraints and local name samples, and re-check civic test data before locker kiosk load tests. Retain the municipal audit note so later testers can see why the decision protects public service quality and fairness. Retain the municipal audit note so later testers can see why the decision protects public service quality and fairness.