Does the “backfire effect” exist—and does it matter for factcheckers?