A recycling-center scale-house tester asks an LLM for cases covering posted surcharge rules. The reply confidently includes a full pack for a “hazardous-oil surcharge” that never appears in the posted rules. Which GenAI defect does this best illustrate?
Select an answer to reveal the explanation.
Short Explanation
Think of a smooth storyteller who invents a plot twist the script never had. The model sounded sure, but the surcharge was not in the rules—so the pack is fluent and still ungrounded.
Full Explanation
When a recycling-center scale-house tester asks an LLM for cases covering posted surcharge rules and the reply confidently includes a full pack for a “hazardous-oil surcharge” that never appears in the posted rules, the GenAI defect is a hallucination that produces fluent but ungrounded testware. Civic fee enforcement must track adopted rules only; invented surcharges can unjustly burden residents if promoted. Labeling it a reasoning error that misorders cause and effect in a fee calculation fails when the core issue is fabricated content absent from the basis, not merely mis-ordered real factors. Calling it bias that under-represents non-English names fails because the defect is invented policy, not skewed demographics. Calling it non-determinism from sampling fails because the distinctive harm is ungrounded substance, even if sampling enabled it. Exam caveat: hallucination = fluent ungrounded artefacts relative to the test basis. Operational check: cross-check every generated surcharge against the posted civic board, delete oil-surcharge cases, constrain prompts to listed rules, and require steward sign-off before scale-house packs enter regression. Retain the municipal audit note so later testers can see why the decision protects public service quality and fairness.