3 ms·

Your HN example makes me think: if I'm talking to my spouse or close friends, we obviously don't have a moderation policy, we know each other well and share val

Your HN example makes me think: if I'm talking to my spouse or close friends, we obviously don't have a moderation policy, we know each other well and share values well enough that any debates we may have are (almost always) focused on substance and not on conduct. In political discourse, and in debates on big platforms like Twitter, it's the opposite- most of the discussion is about people's or groups' conduct and substance takes a back seat. Because a heterogeneous group with different values is involved. So for social media and online forums, the question is, how big and diverse can the audience get while still supporting civil, substance focused discussion? HN does a pretty good job, and also has some obvious biases, scared cows, and weak spots. Online newspaper article comments probably have one of the lowest quality of discourse for a given participant size. What forum is best, I'm not sure, but its instructive to look at it this way, because it reflects politics generally, if we want to address real issues while maximizing participation.

groby_b 6y ago
I have a sneaking suspicion that you can safely support tremendous community size if you don't have enragement-driven ordering of the feed. (I.e. strict chronological ordering only) We tend to pick our online friends according to mostly shared values, the same as real-life friends. Enragement ML then proceeds to amplify the differences instead of letting you experience that you're mostly similar, radicalizing both partners in the long term. A chronological order minimizes that experience simply by the probabilities - if most of your values overlap for everybody in the feed, the chances of you seeing a large chunk you do not want to see is small. You're occasionally reminded that you don't always agree, but it's in a context of lots of other posts that don't violate your values. (Part two of the problem are searchable feeds, where groups of people just wait to keyword-snipe)

groby_b 6y ago

I have a sneaking suspicion that you can safely support tremendous community size if you don't have enragement-driven ordering of the feed. (I.e. strict chronological ordering only) We tend to pick our online friends according to mostly shared values, the same as real-life friends. Enragement ML then proceeds to amplify the differences instead of letting you experience that you're mostly similar, radicalizing both partners in the long term. A chronological order minimizes that experience simply by the probabilities - if most of your values overlap for everybody in the feed, the chances of you seeing a large chunk you do not want to see is small. You're occasionally reminded that you don't always agree, but it's in a context of lots of other posts that don't violate your values. (Part two of the problem are searchable feeds, where groups of people just wait to keyword-snipe)