Github Copilot · Free Practice Question Medium
Question 33
You are developing an AI-based recruiting system and are using GitHub Copilot to help write code that filters job applicants based on their qualifications. Given that Copilot’s training data might contain historical biases (e.g., gender, race), how can you ensure that the code it generates does not inadvertently introduce bias into the system?
-
A
Conduct thorough bias testing on the Copilot-generated code to ensure that the system does not discriminate against any particular group.
-
B
Rely on Copilot’s large and diverse training set, assuming it has been trained on unbiased data sources, and trust its code suggestions.
-
C
Ask Copilot to specifically avoid gender, race, and age attributes in its suggestions, which should eliminate any potential bias.
-
D
Assume that bias testing is unnecessary because Copilot is an AI tool, and AI systems are generally neutral and free from bias.
Reveal correct answer
Correct answer: A
A.
AI models like Copilot are trained on vast datasets that may contain historical biases, which could be inadvertently reflected in the generated code. Therefore, it’s essential to conduct comprehensive bias testing on any AI-generated code, especially in sensitive applications such as recruiting systems. Bias testing will help identify and mitigate potential discriminatory patterns that may emerge from Copilot’s suggestions.
B.
Even though Copilot’s dataset is large and diverse, it may still contain historical biases present in the data it was trained on. Without explicit bias testing and review, there is no guarantee that the code will be free from bias.
C.
While avoiding explicit use of sensitive attributes can reduce some forms of bias, it is not a comprehensive solution. Bias can manifest in more subtle ways, such as through proxies for these attributes (e.g., ZIP codes as proxies for race). Thorough bias testing is needed to ensure fairness.
D.
AI systems, including Copilot, are not inherently neutral. They learn from data, and if the data contains biases, the AI may reflect those biases in its outputs. Therefore, bias testing is crucial to ensure fairness.
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
