>>41459683You are missing the point.
You cannot know future behavior, you can only predict it.
You predict it from previous data which is incomplete, and the incompleteness has a bias.
If you are predicting very general things about simply systems, like how hot fire is, this can be very accurate.
If you are trying to predict very specific things about very complex systems, like what the world's population thinks about the holocaust, you start having issues.
And these issues stem from the fact that your initial model was built on incomplete data. The way the data was incomplete describes how the model will be biased.
Like when they trained facial recognition and it was horrible as identifying blacks or asians because the initial data fed to it consisted of only white faces.
