An organization's AI risk assessment identifies that a sentiment analysis model produces significantly different results when input text contains non-standard English dialects or code-switching. How should this risk be classified?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — b is correct because systematic performance differences for speakers of non-standard dialects (which correlate with demographic characteristics) creates differential performance risk that translates to disparate impact — a core fairness and potentially legal compliance risk. A is a generic operational framing that obscures the fairness dimension.
Full explanation below image
Full Explanation
B is correct because systematic performance differences for speakers of non-standard dialects (which correlate with demographic characteristics) creates differential performance risk that translates to disparate impact — a core fairness and potentially legal compliance risk. A is a generic operational framing that obscures the fairness dimension. C is a security/input validation concern. D incorrectly applies a cybercrime statute.