21-Mitigating algorithmic bias Flashcards
What is out-group homogeneity bias?
Humans tend to perceive out-group members as less nuanced than in-group members
What is correlation fallacy?
Humans mistake correlation with causality
What is historical bias?
A randomly sampled dataset reflects the world as it was including existing biases which should not be carried forward
What is representation/reporting bias?
Datasets do not faithfully represent the whole population
What is measurement bias?
Bias within the dataset, due to:
- Noisy measurement
- Noisy proxy
- Oversimplification of quality of interest
What are biased loss functions?
Blind to certain types of errors
What is evaluation bias?
Bias where the test set is not representative of the target population
What is deployment bias?
Use of systems in ways that they weren’t intended to be used
What is the machine learning pipelien?
World -> measure -> data
data -> learn -> model
model -> action -> individual
individual -> feedback -> model
What are sensitive attributes?
Attributes that we don’t want to decision on
What is fairness through unawareness?
Hide all sensitive features from classifier
