Modelling default and likelihood reasoning as probabalistic reasoning