Our biases reflect statistical relationships that people observe. There is a body of evidence that suggest that these biases are rather accurate. I don't see why we would want to remove them from our models.