Data points on themargins of distributions of human data tend to correspond to marginalizedpeople . Popularity and homogenizing biases have the effect of furthermarginalizing the already marginal people . This source of bias warrants serious attention given the ubiquity of algorithmic decision-making. This is not merely biases in the statistical sense; thesestatistical biases can cause discriminatory outcomes. These are not just biases in statistical sense, they can cause discrimination in the way of learning from user responses to documents that the algorithmrecommended. They are not merely biased in a statistical sense. They can create a selection bias in the course of learning to learn from users responses to the algorithm. It is not just the result of the algorithm’s selection bias.

Author(s) : Catherine Stinson

Links : PDF - Abstract

Code :
Coursera

Keywords : bias - biases - statistical - sense - responses -

Leave a Reply

Your email address will not be published. Required fields are marked *