PRODUCTION SCENARIO
A hospital risk-scoring model was tested only on complete records. When records arrive with lab fields missing, it still returns high-confidence scores instead of flagging the input.
Which responsible AI principle is most directly at stake?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Reliability and safety
Reliability and safety requires a system to 'function as designed, respond safely to unexpected conditions'; its handling of different conditions 'reflect the range of situations developers anticipated during design and testing'. Confident scores on incomplete records show that condition was never tested. Accountability is who can overturn the result, transparency is explaining it, and privacy and security is protecting the record.