Removing algorithmic discrimination (with minimal individual error)

We address for the first time problem of correcting group discriminations within a score function, while minimizing individual error. Each is described by probability density function on set profiles. solve analytically in case two populations, with uniform bonus-malus zones where each population majority. then general n entanglement populations does not allow similar analytical solution. show that an approximate solution arbitrarily high level precision can be computed linear programming. Finally, we reverse error should go beyond certain value and seek to minimize discrimination.