Which are mitigation strategies for data scarcity?

• A. Augmentation, collecting missing data, generating synthetic data, or imputing missing data.
• B. Only collecting more data.
• C. Relying solely on existing data.
• D. Ignoring missing data.

Answer: (A)

Data scarcity in AI training leads to models that don’t generalize well because they haven’t seen enough variation. The best way to mitigate this is by combining several strategies: augmenting existing data to create diverse, realistic variations; collecting missing data when feasible to expand the real-world sample; generating synthetic data to cover rare or unseen scenarios; and imputing missing values to make use of incomplete records. Together, these methods increase the effective sample size and improve coverage of different situations, leading to more robust models. Relying only on the existing data and ignoring missing data won't address scarcity and can introduce bias, while collecting data alone may be too slow or expensive to close all gaps.