Difference between revisions of "Model Bias"
From Open Risk Manual
Wiki admin (talk | contribs) |
(No difference)
|
Latest revision as of 15:41, 4 September 2020
Contents
Definition
Model Bias (also Algorithmic Bias) denotes the systematic and repeatable error in a Risk Model that creates outcomes that are statistically at odds with the system, population or behavior that is being modeled.
Examples
Unfair outcomes include privileging or excluding arbitrary groups of users / or clients over others.
Causes
Bias can emerge due to many factors:
- biases originating in the availability of data
- biases embedded in the data processing pipeline (extraction, cleaning, transformation)
- biases embedded in the model development process (including selection of algorithm)
- biases induced through the particulars of model implementation