It's not like NTs are so great at it. A famous group of experiments demonstrated that with a person asking for directions. In mid-encounter, they were briefly separated by two workmen carrying a big piece of plywood. When they rejoined each other, the directions-asker was a different person.
The more closely the swapped people resembled each other, and the more there was diversity between the directions-asker and the subject of the experiment (age, race, gender) the less likely it was that the person would notice they were not talking to a different person than when the conversation started.
The more closely the swapped people resembled each other, and the more there was diversity between the directions-asker and the subject of the experiment (age, race, gender) the less likely it was that the person would notice they were not talking to a different person than when the conversation started.