In Tokyo, a sustained conversation even at barely audible levels would be discouraged by social comparison - others don't so I shouldn't. As a result I can follow a typical conversation between seemingly oblivious native English speakers from half-way down a carriage.

I don't know how/why but those conversations are a lot quieter than the video chats.

Better audio quality plus non-visual cues?

People yell when they talk to their phone.

Only when using the handset, and on connections without sidetone. With speakerphone style conversations there would be no perceived need to yell.