There are three people in the room helping you out with an estimate. You discuss the next feature, then ask for everyone's opinion. A guy and a girl think it will be two days. The other guy thinks it will be five.
Alright, then (2 + 2 + 5) / 3 = 3 days. You average the estimates out, write down the result and move on. As easy as that.
Except that's an awful idea.
Averaging estimates is even worse than picking the lowest or the highest estimate. When taking a side, at least you are giving credit to someone, but by using the average, you are forsaking and distorting everyone's opinion.
What should you do? Start a discussion. Ask why it should take five days. Ask why it could be as easy as two days. Find out what's making different people give different estimates. The conversation will likely float hidden requirements or misunderstandings about the work. That's what you need to happen before moving on.