Overall, the difficult part is being able to define political orientations. Without that you cannot really proceed with network analysis.

What does the online political landscape look like?
Network nodes defined as YouTube content creators. Links between them are the number of guest appearances from other creators. Size of nodes can be either number of views or number of subscribers
Problems with this: the network may be very sparse, finding all the links correctly might be difficult, guest appearances aren’t the best way to describe ideological overlap/audience overlap!
Also, bread tube is less dominant on YouTube, and a lot of them may be streamers on Twitch or may use YouTube for video-essay purposes
How do right wing and left wing elements differ in terms of size, organisation, connectivity?
First, you would have to separate the networks into left/right based on manual inputs, e.g. how they self identify. Then you can use network metrics (e.g. rich club, assortativity, diameter, etc.) to describe the networks
Is there a red-brown alliance?
Not sure how, but need to look into clustering/community detection to see if there is an overlap between ‘far left’ and ‘far right’ channels. This requires a method for quantifying political orientation, and also you don’t want to capture ‘centrists’ in this
Are there nodes that have the effect of ‘radicalising’ viewers?
If the network is directed, can you find nodes that have the effect of ‘moving’ viewers from the centre to the left/right?
A way you can do this: if you know which nodes you are interested in, e.g. Jordan Peterson and you have different networks, e.g. left and right. You can find the shortest path from JP to all the nodes in left/right domain and them average it! So you can find the average path from JP to each network respectively? Not bad huh? But how does shortest path work for directed networks? Should you be treating the network as undirected?
What types of communities exist?
Can you see if different political representations, e.g. libertarian socialist, right libertarian, soc dems exist in different communities?
Can you use the network information to encode political orientation similar to that of 8values or the political compass?
The idea is to generate data from the network, then using matrix factorisation methods, PCA or alternatives encode that information into 2D and 4D, and try to interpret the axes (e.g. economic/authority for political compass, and the 4 for 8 values)
The problem with this project is that you need o get started in order to be able to proceed…