News.YC statistics - community cohesion

http://clipboarded.blogspot.com/2008/02/newsyc-statistics.html

by matstc • 18 years ago

34 34 18 years ago

clipboarded.blogspot.com

Alex39174• 18 years ago

It's not entirely clear to me that news.yc would be a better place without all the submissions and comments that get only one or two upvotes. For example, I'm still convinced that idiagram's wicked problem solving visualization was one of the intellectually interesting (and beautiful) news.yc submissions ever, even though I'm apparently the only one who voted for it:

http://www.idiagram.com/CP/cpprocess.html

Looking through my own submissions I see that the more intellectual stuff I've posted, while it's done well, has never gotten as many votes as the pithy one-liners.

whacked_new2• 18 years ago

Just looked (again) at that link. I recall this page and recalling myself not getting far through it. On my revisit I still couldn't read it through entirely. It looks like one of those handouts that presents an unspecific problem, to which it throws a bunch of unspecific waypoints that lead to an unspecific solution.

I suspect you see something that I didn't; how do you approach reading this?

Also, between reading and coming back, you have been downvoted once. I have no idea what is downvote-worthy in your post... sometimes I just don't know what I am missing -- or if there is anything to miss in the first place.

ratsbane2• 18 years ago

I've wondered about that too - what motivates people to vote something up or down? Is voting up more a measure of interest, agreement, surprise, appreciation...? Would you vote down because you don't agree with a comment or because you think it's irrelevant, self-serving, unoriginal, or poorly-stated? (I didn't see anything I thought worthy of downvote on the parent comment either; I vote it up as thought-provoking even though I didn't quite get the point of the thing he linked to.)

Alex39174• 18 years ago

"How do you approach reading this?"

Normally people solve problems using some variant of the rational decision making problem. That is, define the problem, define assessment criteria, brainstorm solutions, choose solution, implement solution, measure solution against assessment criteria.

A wicked problem, on the other hand, is a problem that can't be clearly defined and has no clear answer. For example, what is the best way to design a car? If you maximize mpg then the car isn't as safe. But if you maximize safety then the car is heavier so the mpg isn't as good. And both safety and mpg affect style.

The idea of the idiagram thing is that while the rational decision making model may be the best way to solve well-bounded problems, it isn't a very good way to solve wicked problems.

This is roughly where the idiagram page leaves off, but it footnotes the book Wicked Problems and Social Complexity, the first chapter of which is available online. This explains dialog mapping, which is a method of discussing wicked problems in group meetings using mindmaps.

Why is this important?

If you've been to a lot of discussion forums on the web, you probably realize that forums with nested comments tend to have more rational discussion than forums with flat comments. After all, it's just easier to reply to specific points of an argument with nested comments than it is to fisk one with flat comments. You can go back and forth on one point, and then go back up in the discussion later and talk about a completely different point raised earlier.

Trying to solve wicked problems in real life meetings without dialog mapping is sort of the equivalent of an flat discussion on the web. You are sitting around and someone comes up with a complicated proposal. The next person responds to whatever they see in it. And then the other points in the original comment become more or less buried. However if you mindmap all of the points of a person's dialog, then people are able to use the visualization to jump back and forth during the discussion and gain perspective on what's important and what's less important. This ultimately means better solutions to wicked problems in less time, with more focus on the problem and less focus on the personalities involved.

The idiagram then is a way of visualization the final output of the brainstorming via dialog mapping process so that all involved can be on the same page when in comes to understanding the solution and each person's responsibilities and how success will be measured. (The whole submission is kind of recursive, because it's a picture about the four steps of the WP solving process, and making a picture is step three of four.)

Anyway, that's what I see in the submission. If you're the kind of person who sits are and thinks about how you can have brainstorming sessions in real life with the same abstract ideas and rational thinking that often arise online (because of the asynchronousness and nestedness of the internet), then this is kind of a cool thing.

DanielBMarkham• 18 years ago

Agreed.

Let's say you post an amazingly funny technology joke and picture each day, and a friend posts an amazing, in-depth discussion of something that most of the readers would want to consume.

Who do you think is going to score all the points? Certainly not your friend with the in-depth articles.

So does that mean the the community is more represented by people who post quick, funny barbs? Are we a group of late-night technology comedians?

I understand the joke analogy is at the extreme, but you can make the same observation comparing, say, 3-paragraph gossipy blog entries with 50-paragraph in-depth articles. Same principle applies.

bayareaguy2• 18 years ago

Funny you should mention that one - it happens to be something I regularly forward to people who like to snicker at bad uses of graphics. One guy I sent it to couldn't stop laughing for an hour.

Alex39174• 18 years ago

What don't you like about it? It's obviously not something you'd put on a powerpoint slide, but given enough time to sit down and think about it I think it's quite insightful.

thaumaturgy• 18 years ago

Awesome, thanks for that link. Occasionally I find myself staring down the barrel of a particularly thorny problem; the initial responses and challenges to the problem listed in the diagram are spot-on. My own internal approach to such problems pretty closely mimics what's in the diagram.

It's nice to have something that describes that whole process -- faced with a toughy problem? This is what to expect, and lest you forget, here's how you can approach it.

aston2• 18 years ago

Cool.

I'd be interested in seeing similar analysis on a per-comment basis.

aston2• 18 years ago

Also, pg could probably run this on the complete data set, which might better illustrate what's happening with the more prolific members of the community.

ratsbane2• 18 years ago

Nice analysis. Does the complete data include timestamps on up/down votes? The ratio of up to down votes might show how controversial is a posting while the rate of rise/fall might show... whether the posting was submitted at an opportune time?

Or what about looking at voting habits? What's the average/min/max/etc up/down voting ratio?

If you showed to what degree showing more statistics, particularly about new postings or first-page postings would change behavior?

Spyckie• 18 years ago

Thanks for posting this interesting data set, but after thinking about it for some time, I don't think this data offers any insights into yc.news. The leader to user ratio is too high, which skews all the numbers so that nothing really meaningful can be extracted. It's also hard to spot any trends from the point system alone - user identity, time of post relative to other news, and the controversy of the post content can play important factors in the post's point value, and all of these don't necessarily have any correlation with the user who posted it. I think in order to figure out something about the community, someone will have to get their hands dirty and actually look at post content.

matstc4• 18 years ago

There is a lot of talk about what should be done in order to keep the community as it is, about what is a relevant post and what's not. But we have no metric to measure the cohesion, except for 'pleasure'. The analysis in the post is of course incomplete but I believe there is value in a metric that could tell us whether or not we are being steered away from the initial community.

cawel• 18 years ago

Those are nice stats, and I do think we can get some interesting characteristics about our community. I thought it is quite powerful to point to paul, palish, sharpshoot, or pg, and say that they are the community, since they have the best points per post measure. They are nothing less than the best representatives of our community (with regards to posts, not comments).

I'd like to think that the factors you mentioned (traffic, time of the day, controversy of the post) iron out statistically (since all posts are affected by those, there is an argument to ignore their respective influence).

This initiative makes me wish it is only the beginning: I want some more stats! For example:

- It would be interesting to see what's the points per post (PPP) of the leaders, compared to the PPP for most of the others (I agree with you that in the dataset, the leader to user ratio is too high). Maybe one could have a plotted distribution of PPP's across all users.

- I know HN has 8000 uniques/day, what is the proportion of those that are posting? e.g. how many of those users with less than 5 posts?

- also interesting is the proportion of points between posts and comments, per user. Different profiles: some people discuss a lot, some people post a lot, as it was often repeated in discussions. But what's the proportion of those 2 groups?

ivankirigin• 18 years ago

Interesting. I post lots of things that I find obscure and don't expect others to upmod.

I think part of the reason that news.yc is interesting is that there are a number of non-overlapping interest areas within a tight niche. Think about that. Diversity within a narrow field. I like that.

manvsmachine• 18 years ago

As a relatively newer member here, it would be interesting to know how these numbers came about over time; i.e, did the more prolific members start out with a bang, or did they gradually build up steam to reach where they are now?

DarrenStuart• 18 years ago

nice little bit of hacking, one thing I would of done is left pg out, I think he is messing your data up. Fans being fans his karma is off the chart.

matstc4• 18 years ago

Yeah his approval rating is ridiculous, above 13 points per post on his last 180 posts. But that's just one of many things messing up the statistics. It's only good if you want to have a general idea.

shafqat• 18 years ago

Very cool. Was fun trying to find my username on your charts (although I wish it could have been a bit higher up the points per post axis!).

edw5192• 18 years ago

Fantastic! A digitally generated picture is worth 2^10 words!

You've certainly embodied the philosophy of Edward Tufte:

http://www.amazon.com/s/ref=nb_ss_gw/102-8193666-1218533?url...

There must be many cofactors that are hard to capture in an analysis like this. There's the "binaryness" of making it to the front page. There's also "time of day". A very interesting submission posted at a time when those who would be interested are not online may never gain enough traction to make it to the first page. A time-of-day analysis would certainly be interesting.

matstc4• 18 years ago

I'm not sure if you're mocking me or giving me too much credit here.. Really I was hoping to be more rigorous and go more in depth but at some point it just felt silly to hammer a website, scrape the page and never have access to the whole data. Doing this got me thinking about issues like "time of day" though as there must be a way to take into account the rate of new posts and make old posts more 'sticky' when the rate accelerates, in the morning for instance.

I feel like starting my own social news website just to have a play with different algorithms!

edw5192• 18 years ago

"I'm not sure if you're mocking me or giving me too much credit here."

Neither, I'm sure.

(Funny. lately when I get sarcastic here, people think I'm serious, and when I'm serious, people think I'm joking. I guess I need a little more practice writing to strangers.)

Scatter diagrams are one of the greatest underrated ways of conveying information in a heartbeat. You've done a great job of it.

You've also opened up a can of worms. There are so many ways to look at this data, you can keep yourself busy for quite a while. I suspect that won't be a problem, considering how much you already care about this. Keep up the good work and keep us posted.

ldambra• 18 years ago

This kind of topic worries me because from my experience when a community starts worrying about "cohesion", "trolls" and like, this is usually the first sign of it's fall down.

The real problem is what most of us would call a fall down today, that will be called tommorow (and interpreted as) a rise up from other people (thus the extreme difficulty to create efficient moderating systems based on quantitative data). You know your community is irreversibly "corrupted" when this kind of people becomes the majority, and this usually happen surprisingly quickly with the faster growth of visitors.

For some reason I have some hope for HN because of PG being behind it, he kinda knows what he wants and certainly knows what he doesn't want. Still, this might be one of the thoughest challenge he'll be faced with if HN keeps growing. I've known no community that escaped this syndrom yet.

npk• 18 years ago

These plots are begging to be placed on a log-log axis!

News.YC statistics - community cohesion

Related Stories