Exam AWS Certified AI Practitioner AIF-C01 topic 1 question 57 discussion

B: Sampling bias Explanation: Sampling bias occurs when the training data used for an ML model is not representative of the real-world population. In this case, the model disproportionately flags members of a specific ethnic group, likely because the training dataset was not balanced or representative of all groups. This leads to skewed predictions that unfairly target certain populations.

Sampling bias: occurs when the data used to train the ML model is not representative of the overall population, leading to the model performing poorly on certain groups, like in this case where the model is disproportionately flagging people from a specific ethnic group. Why the other options are not correct: Measurement bias: This refers to errors in the way data is collected or measured, which isn't directly related to the ethnic group bias in this scenario. Observer bias: This happens when a human observer's personal biases influence their interpretation of data, which isn't applicable here as the model is making the evaluations automatically. Confirmation bias: This refers to the tendency to seek out information that confirms existing beliefs, which isn't relevant to the training data used to develop the ML model.

B. Sampling bias Explanation: Sampling bias occurs when the data used to train a model does not accurately represent the diversity of the population or real-world scenarios the model will encounter. In this case, if the training data for the security camera footage had an overrepresentation or underrepresentation of certain ethnic groups, the model may disproportionately flag members of that group as potential theft suspects. This leads to biased predictions due to imbalanced or unrepresentative training data.

A. Measurement bias Measurement bias occurs when the measurement process itself is flawed, leading to systematic errors. In this case, the model is likely biased due to the way it's trained on data that may not be representative of the entire population. This can lead to the model incorrectly associating certain characteristics with criminal behavior, particularly for individuals from underrepresented groups.

B - Sampling bias occurs when the data used to train a model is not representative of the population or real-world scenarios it's meant to analyze. This leads to skewed results that favor or disfavor certain groups.