Artificial Intelligence & Machine Learning
Bias
Definition
In the context of machine learning, bias refers to systematic errors in the model that result from incorrect assumptions in the learning algorithm. More broadly, it can also refer to the way a model reflects the societal biases present in its training data.
Why It Matters
AI bias is a major ethical concern. If a model is trained on biased data, it will learn and often amplify those biases, leading to unfair or discriminatory outcomes when deployed in the real world.
Contextual Example
If a hiring model is trained on historical data from a company that predominantly hired men, the model may learn to associate male candidates with successful hires, leading it to unfairly discriminate against female candidates.
Common Misunderstandings
- This "societal bias" is different from the statistical "bias" in the bias-variance tradeoff, though they are related.
- Auditing models for bias and ensuring fairness are critical and active areas of AI research.