Which of the following approaches BEST helps reduce model bias?

The correct answer is A. Ensuring diversity in training data sources. Model bias primarily originates from non-representative or skewed training data. Ensuring diversity in training data sources directly addresses the root cause by exposing the model to a broader range of real-world distributions, demographics, and scenarios. A more complex archite

AI Security Design and Implementation

Question

Options

AEnsuring diversity in training data sources
BUtilizing a more complex architecture
CDecreasing frequency of model updates
DIncreasing the number of labels per instance

How the community answered

(18 responses)

A
89% (16)
B
6% (1)
D
6% (1)

Explanation

Model bias primarily originates from non-representative or skewed training data. Ensuring diversity in training data sources directly addresses the root cause by exposing the model to a broader range of real-world distributions, demographics, and scenarios. A more complex architecture (B) does not fix data-level bias - it may even amplify it. Decreasing update frequency (C) can cause the model to drift from current reality, potentially worsening bias. Increasing labels per instance (D) improves annotation granularity but does not resolve underlying distributional imbalances in the dataset.

Topics

#Model Bias Mitigation#Training Data Diversity#Responsible AI#AI Ethics

Community Discussion

No community discussion yet for this question.

Full AAISM Practice