I think that it's actually not an IRL vs Internet discussion dichotomy FWIW, that oversimplifies things a bit too much. For example, I think sympathy and empathy increases as you go from public interaction in major social platform -> 1-on-1 direct message -> audio call -> video call -> IRL private conversation. Many reasons really. Public interactions have public scrutiny, it often is more of a performance than an authentic social interaction. Text conversations are low bandwidth. All internet conversation carries the burden that things you say may be recorded and used against you. And of course, seeing and hearing a human on the other end just naturally will increase empathy.

I hear what you're saying, but I think you are looking at this wrong. I'm not suggesting that IRL social interactions are perfect and Internet ones are hopelessly broken. We have plenty of history to show how IRL social interactions can break down, in small groups or big. What I'm saying is that IRL interactions are better on average than Internet interactions, largely due to the modality and venue that most Internet interactions occur.

Edit: also, it is worth noting that there are definitely robust studies that can back up some of these ideas, which may be a good data point to add into this discussion. That said, I am honestly too lazy to go cite sources right now, to be completely honest with you. (And I know HN is rightly a bit skeptical of psychology studies, since you can pretty much find something to validate anything you want, but there are some actually good and interesting ones.)