Jonny
null
- Joined
- Sep 8, 2009
- Messages
- 3,134
- MBTI Type
- FREE
I've been curious about the relatively recent surge in tribalism since the 2016 election. It is clear to many of us who frequent the politics sub-forum, both as contributors and readers, that here at TypoC we're not immune to this phenomenon. But, unlike our daily interactions in the "real world", here at TypoC we have a wealth of data at our fingertips to shed light on this very issue. By analyzing meta-data on the interactions between forum members, we can better see what lies under the surface.
So, over the past few days I've been gathering data from the politics sub-forum in a similar(ish) fashion to what was done here: What your posting habits say about you!
In particular, I've focused on 5 major threads:
As of today, I haven't really done much analysis, and I've only gathered very simplistic data like poster names and "likes". However, even this information has been illuminating. Within these threads are just over 16,000 posts by close to 300 members, with a 6,700 likes spread between about 4,500 posts.
As a simple way to visualize these factions, I decided to extract groups of Likers and Posters. The Likers are those members who liked the most posts within the data set, concentrated between the fewest number of members who posted; for example, if you liked at least 50 different posts, and the members you liked had an average of about 3 likes from you each, you were included in this group. The Posters are those members with the most likes in the data set, concentrated between the fewest number of members who liked them; in other words, if you had at least 30 likes and you received roughly 4 likes from each member who liked you, you were included in this group.
The reason for including the concentration parameter is to ignore those members who have like 100 likes from 100 different users, since they would likely not represent a "faction" of the forum.
From this information, I created a matrix for the numbers of likes given by each of the Likers identified above to each of the Posters identified above. I then, very roughly, normalized these numbers to account for the fact that some Likers liked more than others, and some Posters were liked more than others. Below is what came out (after a bit of manual sorting). White colors denote no or very few likes, while green and blue represent increasingly significant concentrations of likes. Squares denoted with an X are for those members who were included in both groups; they're colored blue because I assume (maybe naively, haha) that everyone likes themselves.
NOTE, IF YOU ARE INCLUDED IN THIS MATRIX I AM IN NO WAY SUGGESTING YOU ARE IN A TRIBE. YOU WERE SIMPLY ONE OF THE MEMBERS IDENTIFIED VIA THE CRITERIA MENTIONED ABOVE. I DIDN'T PICK ANY MEMBER EXPLICITLY.
Keep in mind that I didn't choose anyone on this list. I set the criteria based on anonymized data, and organized the matrix before knowing who was on it. Then, I converted the anonymous key to member names, and voila! Also, these numbers do not represent the number of likes; they are normalized! So, it isn't the case that Xann liked 0 of my posts; I'm not sure of the exact number but it's at least 1. Rather, based on the average number of likes received and given by everyone in this matrix, I we got a score of 0. Blank squares, however, do denote 0 likes.
Obviously this analysis is somewhat simplistic; I'll continue to research the best ways to parse and present this information. In addition, I'm curious about analyzing additional data; for example, using a word analyzer to assess the tone/aggression in a post, and flagging users who engage in back-and-forth posting (denoted, for example, by quoting one-another).
Interesting stuff..
@Jaguar
@Magic Poriferan
@bechimo
@geedoenfj
@Floki
@Totenkindly
@jcloudz
@Hard
@Ivy
@Lark
@ChocolateMoose123
@Poki
@asynartetic
@SearchingforPeace
@SpankyMcFly
@ZNP-TBA
@Stephano
@Xann
@Tellenbach
@S16M4
@DiscoBiscuit
@YUI
@Virtual ghost
@Jonny
@ceecee
@Nicodemus
So, over the past few days I've been gathering data from the politics sub-forum in a similar(ish) fashion to what was done here: What your posting habits say about you!
In particular, I've focused on 5 major threads:
- Trump Administration
- United States of America - General Thread
- TRUMP 2016!!!!
- Europe - general thread
- 3rd wave feminism
As of today, I haven't really done much analysis, and I've only gathered very simplistic data like poster names and "likes". However, even this information has been illuminating. Within these threads are just over 16,000 posts by close to 300 members, with a 6,700 likes spread between about 4,500 posts.
As a simple way to visualize these factions, I decided to extract groups of Likers and Posters. The Likers are those members who liked the most posts within the data set, concentrated between the fewest number of members who posted; for example, if you liked at least 50 different posts, and the members you liked had an average of about 3 likes from you each, you were included in this group. The Posters are those members with the most likes in the data set, concentrated between the fewest number of members who liked them; in other words, if you had at least 30 likes and you received roughly 4 likes from each member who liked you, you were included in this group.
The reason for including the concentration parameter is to ignore those members who have like 100 likes from 100 different users, since they would likely not represent a "faction" of the forum.
From this information, I created a matrix for the numbers of likes given by each of the Likers identified above to each of the Posters identified above. I then, very roughly, normalized these numbers to account for the fact that some Likers liked more than others, and some Posters were liked more than others. Below is what came out (after a bit of manual sorting). White colors denote no or very few likes, while green and blue represent increasingly significant concentrations of likes. Squares denoted with an X are for those members who were included in both groups; they're colored blue because I assume (maybe naively, haha) that everyone likes themselves.
NOTE, IF YOU ARE INCLUDED IN THIS MATRIX I AM IN NO WAY SUGGESTING YOU ARE IN A TRIBE. YOU WERE SIMPLY ONE OF THE MEMBERS IDENTIFIED VIA THE CRITERIA MENTIONED ABOVE. I DIDN'T PICK ANY MEMBER EXPLICITLY.
Keep in mind that I didn't choose anyone on this list. I set the criteria based on anonymized data, and organized the matrix before knowing who was on it. Then, I converted the anonymous key to member names, and voila! Also, these numbers do not represent the number of likes; they are normalized! So, it isn't the case that Xann liked 0 of my posts; I'm not sure of the exact number but it's at least 1. Rather, based on the average number of likes received and given by everyone in this matrix, I we got a score of 0. Blank squares, however, do denote 0 likes.
Obviously this analysis is somewhat simplistic; I'll continue to research the best ways to parse and present this information. In addition, I'm curious about analyzing additional data; for example, using a word analyzer to assess the tone/aggression in a post, and flagging users who engage in back-and-forth posting (denoted, for example, by quoting one-another).
Interesting stuff..
@Jaguar
@Magic Poriferan
@bechimo
@geedoenfj
@Floki
@Totenkindly
@jcloudz
@Hard
@Ivy
@Lark
@ChocolateMoose123
@Poki
@asynartetic
@SearchingforPeace
@SpankyMcFly
@ZNP-TBA
@Stephano
@Xann
@Tellenbach
@S16M4
@DiscoBiscuit
@YUI
@Virtual ghost
@Jonny
@ceecee
@Nicodemus
