Students' confidence distinguishes accurate from inaccurate model-evaluation trials, with a large printed effect size (d = 1.09), though students remain often confident on inaccurate trials