When in doubt: Using confidence and consensus as ‘summary statistics’ of collective knowledge