In a nutshell: don’t shoot talking heads talking and looking to someone else (not the camera) without a context of a conversation. It disorients the viewer and looks lame.
It is a good approach to switch between 2 shots while shooting a “talking head”:
1) A speaker is watching directly in the camera
2) A speaker is having a conversation with someone behind the camera
Such approach creates rhythm and makes the video visually diverse.
But this good trend has turned into a bad one in just a few years. People began to specifically remove from the frame any details that will help you understand that this is really a conversation between a speaker and someone else, not a crazy speaker who suddenly starts talking past the camera.
How to fix it? Bring back the context of a conversation. Start putting in the frame some parts of the shooting equipment: a lamp with back or side light, which was set specially for the shooting; a microphone; it is possible to take the plan wider, capturing not only the speaker, but also the table with water in front of it. Basically, it can be anything what will help the audience to identify the context as a conversation with someone else.
Of course, you don’t have to worry about that if you have a third type of shots — a face of a journalist asking questions, a black screen with titles, etc. In that case there will be no effect of speaker suddenly speaking past the camera. Previous frames would have already set up the context as a conversation with somebody.