Assessing Multiple Prior Models of Behaviour under Ambiguity