Towards an Inductive Logic Programming Approach for Explaining Black-Box Preference Learning Systems