Survivorship bias refers to the phenomenon of only observing samples that have 'survived' after a selection process while ignoring the large number of eliminated samples, thereby drawing distorted conclusions.