A hospital deploys an AI system to predict patient readmission risk using historical health records. To protect patient privacy, the hospital wants to ensure that individual patients cannot be identified from the data used for training. Which responsible AI principle is most directly relevant to this requirement?
The Privacy and security principle is the correct match because this AI predicts readmission from electronic health records, which are protected health information (PHI). This principle mandates robust access controls, encryption, de-identification, and anonymization techniques to prevent unauthorized re-identification of patients. It directly addresses the need to ensure no individual can be identified from the training data, making it the only option that covers data protection.
Why this answer
The requirement to prevent individual patient identification from training data directly aligns with the privacy and security principle, which mandates data anonymization, de-identification, and access controls. In AI systems, this is implemented through techniques like differential privacy (adding noise to data) or k-anonymity to ensure that outputs cannot be re-identified. The hospital's goal is to protect patient confidentiality, which is the core focus of this principle.
Exam trap
Microsoft often tests the distinction between privacy (data protection) and fairness (bias mitigation), causing candidates to confuse anonymization with equitable outcomes.
Why the other options are wrong
The requirement is about preventing identification of individuals from training data, which directly relates to data privacy and security, not fairness. Fairness addresses bias and equitable treatment across groups, not individual identifiability.
The requirement is specifically about preventing identification of individuals from training data, which directly relates to privacy and security, not to the system's reliability or safety in making predictions.
Inclusiveness focuses on ensuring the AI system works well for diverse user groups, not on protecting individual patient identities from training data.
When would these options actually be correct?
A loan approval AI system uses historical data that contains biased decisions against a certain demographic group. The bank wants to ensure the model does not discriminate. Here, Fairness is the most relevant principle.
An AI system for diagnosing diseases must consistently produce accurate results under varying conditions and not cause harm. If the question asked about ensuring the system performs correctly and safely, reliability and safety would be the correct principle.
A question asks which principle ensures an AI system provides equitable outcomes across different demographic groups, such as race or gender, without bias.
Why candidates pick the wrong answer
Candidates may confuse fairness with privacy because both involve ethical handling of data, but fairness focuses on group bias while privacy focuses on individual data protection.
Candidates may confuse the need to protect data (privacy) with the need to ensure the system works correctly (reliability), especially when the system is used in a high-stakes healthcare setting where safety is a common concern.
Candidates may confuse 'inclusiveness' with data privacy, thinking that protecting patient identities is about including all patients fairly, but it's actually about confidentiality.