A Network of Reddit Users
What can we tell from a network of reddit users, where they connect from being active on the same subreddits?
The network is Dense!
On average users have at least one common subreddit with every 2nd user of the network. This results in almost two million connections among our 2669 analysed reddit users! This is a very dense network, which can be problematic. If a network is too dense, it becomes difficult to extract information from it. Imagine if every user were connected, i.e. had written on the same subreddits - then there would be no point in investigating what type of users write on the same subreddits! To the left the network is drawn, where the dense nature of the network appears, as the nodes are all drawn together. Blue and red nodes (the dots) are Biden and Trump users respectively, and the black edges (lines) are connections.
But if the network is this dense...
is network analysis then suitable?
There are (almost) NO limitations to what we can do with network analysis. Some bright people - M. Ángeles Serrano, Marián Boguñá and Alessandro Vespignani proposes a way of extracting the significant information of a weighted network - the so called Backbone of the network. The main idea is to look individually at each user’s connections and only keep those who provide a significant amount of information, rather than setting a global threshold for all users. For example, if we require users to have at least five subreddits in common before linking them, we might loose very informative links of two users with e.g. three subreddits in common.
The Network
Hubs and central users
For our network the method removes more than a million connections reducing the number of connections with ~97.5%. Most of these are of users where they only have one top-100 subreddit in common (for instance; Ask Reddit), which is a rather non-informative connection, as the majority of all users have this subreddit in common. The resulting network appears as what is called a scale-free network. This type of network is very common among social networks and is characterised by having large hubs, which are very connected nodes compared to all other nodes. This tendency can be seen from what is called a degree distribution, which shows the overall picture of how connected users are. The degree distribution is shown to the left, which reveals that the majority of users only have a few subreddits in common (the first large pillar), whereas a decreasingly number of users have many subreddits in common (the gradually smaller pillars going right). When looking closer at these most connected (or central) users, we see that they typically have comments on:
​
-
The generally popular subreddits, like the top-100 subreddits
-
Political Subreddits
-
News related subreddits.
​
This makes great sense as these type of subreddits are the most frequent in the dataset (see overview). To see which subreddits the three most central users are commenting on, click here.
How does the backbone of the network look like?
To the right we see a drawing of the Backbone, made with a force directed layout algorithm. This type of algorithm works by simulating a physical system where the nodes repulse each other but at the same time attract other connected nodes proportional to their weighted connection (i.e. how many subreddits they have in common). From this we can observe two things
-
First, we see that nodes/users are a lot more spread compared to the previous drawing. This tells us that nodes are pulled towards each other with less force, hence are they not as connected anymore.
-
Second, We see that red and blue nodes tend to group together, which suggest that red nodes have stronger links to each other than to blue nodes and vice versa for blue nodes. This implies that Trump users (red nodes) generally have more subreddits in common with each other and similarly for Biden users! But to what extent does this reflect a polarisation?
So - Do Biden and Trump users write comments on different subreddits?
The short answer is - to some extent!
The longer answer is that within network science, we typically use “modularity” as a measure to what extent nodes in a network groups together in what we call communities. This modularity score is high if nodes in each community are mostly connected to other nodes in the same community. And if we treat Biden and Trump nodes as two separate communities, this yields a modularity score of 0.09. From this score alone we cannot claim a significant community structure - only that users in each community are more connected than some random partition. However, thinking about the nature of the reddit data, it makes sense that communities are not very separated, when we consider that most users typically have some subreddits in common - namely the typical popular ones like the top 100 subreddits. When using other tools which
indicates grouping of users in a network - for instance the louvain algorithm, assortativity with respect to this node property or the force-directed plot presented above - these all point towards some polarisation between users who have commented on the Trump and the Biden subreddit page.
The subreddits which polarise
If we look isolated at the two communities of users, we can extract the most commented subreddits within each, to see what primarily separate them (the image titled unfiltrated to the right). To further emphasise the differences, we’ve also included the filtrated most commented subreddits, where we have removed the 100 most popular subreddits among all users on reddit. This reveals some very interesting things about what contributes to the polarisation of users. Move your mouse over each point below to see!
​
​
-
A lot of Trump users have commented on subreddits related to republican politics.
Where Biden users only have
As the democratic counterpart.
-
​In addition to the above point, a fair part of the Biden users have commented on subreddit pages which are opposed to Donald Trump
-
Biden users are much more active on general politic and news oriented subreddits
compared to Trump users (note the difference in frequencies and rank):
-
Finally, as a funny little “contrast”, Trump users are typically commenting on the subreddit page related to conspiracy theories
Where Biden users are usually more active on the subreddit related to the Coronavirus
​
Filtraded
No. 0 Conservative
No. 8 Tucker Carlson: The Sworn Enemy of Lying, Pom..
No. 16 AskThe_Donald
No. 19 Steven Crowder
​
No. 12. The Ideological Trashcan
most frequent subreddits among each community of users, with the top 100 subreddits removed. For instance, we can see that 473 of the Trump users have a top comment on the subreddit “Conservative”).
Filtrated
most frequent subreddits among each community of users, with the top 100 subreddits removed. For instance, we can see that 473 of the Trump users have a top comment on the subreddit “Conservative”).
Unfiltrated
most frequent subreddits among each community of users. For instance, we can see that 473 of the Trump users have a top comment on the subreddit “Conservative”).
No. 0 Politics
No. 2 All news, US and international
No. 3 World News
​
​
​
No. 4. All news, US and international
No. 5. Politics
No. 8. World News
​
Filtrated
most frequent subreddits among each community of users, with the top 100 subreddits removed. For instance, we can see that 473 of the Trump users have a top comment on the subreddit “Conservative”).
