Is it feasible to ML classify contacts according to whether they're (1) picking a fight; and (2) don't intend to cover the costs; and (3) trying to dump costs saboteurially?

This may be similar to the "classify according to [Graham's Hierarchy of Disagreement]" task?