Some Medieval townsfolk thought witches were poisoning their wells. Witches, of course, were people—often ugly—who were in league with Satan, could do magic, and had an affinity for broomsticks. These villagers wanted to catch the witches so they could punish them. Everyone in the town felt that witches should be punished, so they set up a commission to investigate the matter.
The commission found that little dark-purple creatures were poisoning the wells. The creatures weren’t human. They didn’t have broomsticks. They couldn’t do magic. They weren’t in league with Satan. They were just another sapient species on earth who happened to poison wells.
The commission then split into two factions. Faction one argued that these purple creatures are what we should mean by “witches.” Their argument: “We used to think ‘stars’ were holes in the firmament, but now we know they are massive fusion reactions many light years away. We revised our star concept, why can’t we revise our witch concept? If we continue using the witch concept, none of our laws have to change!” Faction two argued that witches didn’t exist, and that a new word was needed. Their argument: “A huge component of our witch concept is that they are humans and can do magic. These purple creatures are so different from our traditional conception of witches that we should accept that witches don’t exist, and these things should be called ‘Plerps’.”
Unless either conception is internally inconsistent or upon reflection deeply uncompelling, there is no epistemic reason why you should side with one faction over the other. You just have to ask yourself which conception does less violence to your conceptual scheme, and what the political and pragmatic effects of supporting faction 1 or faction 2 would be.