問題文
Removing three personal attributes from the training data would satisfy the enterprise's minimization standard but reduce the model's accuracy. What should the practitioner recommend?
選択肢
- Test whether the accuracy loss is material for the decision before deciding because the standard applies unless the attributes are necessary.
- Keep the attributes because a more accurate model produces better outcomes for the individuals concerned and the minimization standard was not intended to require the enterprise to make worse decisions about the people whose data it holds.
- Keep the attributes in the dataset but restrict who can see them in the training environment and log every access to them for periodic review by the data owner.
- Remove the attributes without testing since the standard is mandatory and the accuracy consequence is a matter for the business to absorb as part of complying with the enterprise's own requirements.