But putting that asside, do you know of attempts to design social networks that won't cause harm?
Yes. Unfortunately Diaspora was just too hard; Google Plus stole a lot of their ideas, they were developing server-side rather than client-side and they weren't nearly monetized enough. This is the problem: a social network should be a public works project, not a private group, because if you attempt to monetize them they turn evil.
Like rd95 I think the attempts made here at Hubski are a step in the right direction, but it'll never be Facebook (or anything like it). At best it'll be a kinder, gentler Reddit which is the best possible thing. There are needs for anonymous networks and named networks and Hubski is the best design I've seen for anonymous.
That Facebook has become "your parents' social network" says a lot about the failings of monetization and anonymity. The rise of SnapChat says more (at least, in principle - in fact, they're lying about anonymity and forgetability). Users WANT a social network that exists for them, not for Zuckerberg and his advertisers, but there haven't been any viable alternatives yet.
Another basic problem is that social network design, right now, is all about engagement, not fulfillment. How many times you come back to the site and mash "like" buttons vs. finding out what your friends are up to. Bragging about your triumphs and burying your failures rather than presenting your honest self to the people that matter to you. These qualities can exist on a social network, but there's no business model that supports a social graph that you interact with as little as possible.
It'll happen, though. Kids are turning away from the invasive, kudzu-like social networks that erupted in the '00s. They just don't have a viable alternative right now.