An evaluation of a forced-choice differential accuracy approach to the measurement of supervisory empathy