Generative AI Leader · Free Practice Question Medium
Question 15
A company develops a generative AI tool to help recruiters screen resumes. After deployment, an audit reveals that the tool disproportionately flags candidates from certain demographic groups as "less suitable," even when their qualifications are comparable. The training data primarily consisted of historical hiring data from an industry known for underrepresentation of these groups.
What is the most likely issue causing this undesirable outcome?
-
A
Poor data accessibility.
-
B
Bias in the training data leading to an unfair AI model.
-
C
Lack of model transparency.
-
D
Insufficient model accountability measures.
Reveal correct answer
Correct answer: B
Explanation
The model's output is reflecting a known imbalance in its training data, a classic example of bias.
INCORRECT: Lack of model transparency.
While a lack of transparency (explainability) can make it harder to diagnose bias, the root cause described in the scenario is the skewed historical data itself, not the inability to interpret the model.INCORRECT: Poor data accessibility.
Data accessibility refers to how easily data can be obtained and used. The issue here is the content and representativeness of the data that was accessed, not the difficulty of accessing it.CORRECT: Bias in the training data leading to an unfair AI model.
Generative AI models learn patterns from the data they are trained on. If the historical hiring data reflects existing societal or industry biases (e.g., certain demographic groups being historically underrepresented or undervalued), the model will learn these biases and replicate them in its resume screening, leading to unfair outcomes.INCORRECT: Insufficient model accountability measures.
Accountability measures are important for addressing issues once they are identified and for assigning responsibility. However, the origin of the problem itself is the biased training data that influenced the model's behavior.
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
