The problem is that current social media software just makes it too easy for extremist bubbles to form. There is no feedback, no discussion, no diffusion, no guiding culture and plenty of false information spread by people who would be unable to publish anything to larger audiences if it wasn't practically for free. To a certain extent, economic systems do filter out people who are challenged at seeking truth. Internet technology has weakened this filter and here we are witnessing attempts at fixing this problem.
But people not only "bubble" themselves, Google and Facebook (everything personalised to one's liking) heavily encourage that, so using the "bubble effect" as justification for removing radical content is really strange.
I was just hinting at what I think is the core problem. If you want to avoid radicalizing bubbles without blocking content, you need to encourage diffusion of ideas, discussion and confrontation. Perhaps, however, radicalized people are just too closed-minded for any such software solution. Then the question is what is worse: An occasional discussion about the decision boundary of what kind of content should be blocked, or dealing with radicalized groups of people more often, potentially risking more terror attacks and also the freedom limiting laws that ensue.
