Modeling and validation of biased human trust