I think that part of it, is that when we are engaging with people in realtime (especially face to face), our emotional driver is behind the wheel, and when there's a pause between responses, our analytical driver has time to grab the wheel, and that's where the "dehumanization" comes from.

That's not always a bad thing. In emotionally-charged situations, that "few seconds of consideration" can help stabilize the interaction.

People claim that it's the lack of consequences, and illusion of safety, but I feel as if it's really the emotional disconnection that does it.