If you throw marbles randomly into a funnel, some of them will go straight in while others will bounce off the bowl bit first before entering after a time delay. Replace the marbles with continuous waves, and the relative delay results in a dip at a certain frequency caused by phase cancellation/destructive interference.

In practice, your pinna (the part of your ear that sticks out of your head) creates this notch at somewhere between 6-12 kHz. Because your pinna isn't perfectly round, the time lag (and thus where and how deep this notch in the treble is) depends on the location of a sound source, which your brain uses for sound localization.

Since headphones are beaming sound right at your pinna, they can potentially acoustically interact with it in a way that creates this notch, as you can see in dummy ear mic measurements like this. IEMs can also be tuned with dips in the treble that artificially simulate this notch.