News for hackers. Stuff that matters.
Re:Division By Three (2006) by dang on Monday April 28, 2014 @02:27AM (#7658500)
I do like the idea of being able to endorse a comment as particularly exceptional. One special and wonderful case of this which happens pretty often is when a person with first-hand knowledge about a story shows up in a thread. (Edit: ran across an example five minutes after I wrote that: https://news.ycombinator.com/item?id=7657553). It's not clear exactly how such a feature would work, though.

About this:

> I feel like part of the reason behind the rise of "dumb and/or mean" is that as HN has gotten larger the community has also attenuated. Commenters become a faceless crowd and the quality of our interactions degrades accordingly

... my main question is how we could test that. It might be true, it might not; hard to know.

Re:Division By Three (2006) by sgentle on Monday April 28, 2014 @07:43AM (#7659296)
Hm, good point.

I suppose there are two parts to my assertion:

1. When you think of yourself as talking to an actual person, you are less likely to be dumb and/or mean.

2. As HN has grown, the degree to which the average commenter considers themselves to be talking to an actual person has decreased.

I was pretty sure I'd be able to find a psychology or sociology study somewhere giving some evidence on (1) or a more general case of (2) for large groups. Wildly, it looks like all of the interest is on how groups reduce your personal identity (deindividuation), not how they reduce the identity of others to you.

So I guess without an appeal to prior research, the best thing would be to develop a measure of "does this comment seem like it's addressed to a real person". You could do that by manually going through and rating them, as (I believe?) you already did with quality metrics.

You could also come up with some heuristics and cross-check them with your own assessment. Here are some I can think of:

* Use of parent's username

* Use of first names

* Use of other usernames

* Use of "you", "your".

* Email addresses, phrases like "contact me", "email me", etc.

* Questions.

* Comment reply chains (ie people replying to a descendent of their own comment)

* "Please", "Thanks" and "Thank you"

I'm not saying that I think comments with this quality are necessarily good, but my prediction is that per (2) the proportion of such comments correlates inversely with the number of active users and per (1) that proportion correlates with a drop in comment quality.

Re:Division By Three (2006) by maxerickson on Monday April 28, 2014 @09:08AM (#7659609)
A lot of times when people use first names or user names they are personalizing a conversation in a context where that doesn't make sense (unfortunately, they are often doing the personalization as part of an attack).

I guess the positive signal still outweighs the negative, but it isn't always a positive signal.