Unravelling multi-agent ranked delegations