The problem is that you are trying to use tools evolved for 12000 years ago1 for problems of the XXI century. Spoiler, it won't work.
I think that back in the day a balanced empathy will make us be helpful with those acquaintances and friends when they are in temporary need. But since we sort of knew everybody1 you know who is in real need because shit happens and who is a careless irresponsible dumbass who got what he/she deserved.
That model doesn't scale, no matter how much of your resources you pour on helping those in needs there will always be more, so at some point is not about "can I help them?" is "do I want to help them?" and that is not a nice place to find ourselves in :( . How we deal with this? I really don't know but I believe that we may evolve a solution, we always did.
Online I think that time and exposure to problems will do it's magic and help us develop new social norms, practices, a sort of "antibody" that will give us the practical wisdom to deal with that too. Is hard to predict since is complex and that is why the solutions need to evolve on its own, but when I see other cases I see people developing practices that counter this behaviors2.
Footnotes
Small communities less than the Dunbar number. That make plausible to get a grasp of who is who all the time. ↩↩2
If you see places with particular and persistent problems like being too cold for example, people develop common wisdom over time + trial and error. Or if you go to the most savage free market environment like dark markets, people also develop unspoken rules that help them avoid common threats like phishing sites, identity theft, etc... If you know that you are looking for you will find a lot more of this examples but as the author of "The design of everyday things" say "Good design is actually a lot harder to notice... because it fit our need so well that the design is invisible" that is why we tend to be oblivious to this social responses to problems. ↩
Footnotes