To take another angle, we could remove humans from the picture entirely. The problem of how to get autonomous agents to work together to accomplish a task is a computer science problem. I don't see a fundamental barrier to solving that problem. Getting humans to cooperate is a special case of that problem, with the extra requirement that you need some domain knowledge in the social sciences to be able to convince people to cooperate. Again, there's no fundamental barrier preventing us from doing that research.
That doesn't mean it's easy. But that's why I'm skeptical when someone claims that the problem is simply unsolvable. Whether intentional or not, people who deny that progress is possible are reinforcing and justifying the status quo.