Fairness in machine learning modeling
To assess fairness, we need to have specific considerations in mind and then use proper metrics to quantify fairness in our models. Table 7.1 provides you with some of the considerations, definitions, and approaches to either evaluate or achieve fairness in machine learning modeling. We will go through the mathematical definitions of demographic parity, equality of odds or equalized odds, and equality of opportunity here as different group fairness definitions. Group fairness definitions ensure the fairness of groups of people with common attributes and characteristics instead of individuals:
|
Topics in Machine Learning Fairness |
Description |
|
Demographic parity |
Ensures predictions are not dependent on a given sensitive attribute, such as ethnicity, sex, or race |
|
Equality of odds |
...
