Model uncertainty and party learning