* Self-moderating: Such conversations require a certain level of decorum, since these are presumably people you interact with every day. The likelihood of calling your desk mate a Nazi, racist, race traitor, etc. to their face are astronomically smaller than when it's cloaked in anonymity, since there may be consequences to your commentary.
* Self-authenticating: It reduces the efficacy of message amplification, since the real world serves as an out-of-band identity authentication mechanism.
* Self-regulating: It's easier, cheaper, and less risky for the hosting site, since they no longer have any enforcement role to play in the discussion itself (beyond the original slant when presenting the information, e.g. by news organizations).
Of course, this doesn't address the problems of sites which are inherently designed to do one-to-many broadcasting and discussion, such as Youtube or Reddit. How would these sites look without comments at all, or with a very strict attestation/vouching system?
Although HN follows a similar model vis-à-vis discussion and broadcasting, the active HN moderation is pretty good, at least good enough to maintain a relatively high-quality discourse and removing very low-quality discussion. However, this is difficult to scale and relatively expensive.