Skip to main content

Machine Bias

There are two lessons I take out of this story:

  1. Promote Free Software, Free as in Free Speech, rather mandate their use in such cases. Even Free Software can also be biased but in case of non free software it can become practically impossible to find their biases and correct them.
  2. We as should selling algorithms as magic, instead of saying there is 60 likelihood that person X will commit the crime again, say according to my system the person X is sixty percent similar to other people who commit the crime again with my rather transparent definition of similarity. Ram and Rahim are brother, as brothers of course they are similar at many levels, Ram has commited a crime does that mean Rahim will do it too?