I agree with you, it is not clear from the test what you're trying to optimize; sorting rotten fruit vs. detecting cancer in xrays are two examples of 'visual examination' where optimal behavior is very different.

So this only makes a good point when your confidence is not well calibrated (rather than consciously settling for lower confidence). I suspect the "impulsive" group Stenovich has in mind are people who are "sure" of the wrong answer, and do not update with feedback.