Been getting obsessed with networks and social graphs recently.
So I built a thing that scrapes and visualizes my entire Twitter social graph.
🐦 twitter.amirbolous.com
Some fascinating insights/learnings about networks, virality, connectivity, and the Twitter-sphere: 1/n
Conversation
Replying to
Research shows many networks (e.g. social networks, Ethereum/bitcoin transaction graph, etc.) are what we call "small-world" networks.
Networks composed of highly dense clusters or neighborhoods with some nodes that connect these different communities.
2
8
We see that for the most part, my Twitter (and probably yours) clusters into demarcated neighborhoods by different topics.
Green ~= Crypto Twitter
Pink/Purple ~= College/New-Grad Tech Twitter
Blue ~= General Tech/Startup Twitter
Black ~= ML Twitter
Orange ~= No-code Twitter
1
5
Within clusters, there's high density, with some people very embedded in the neighborhood. E.g. in Crypto Twitter here we can see all the Paradigm folks , , , popular protocols , @LensProtoco, shitposters , etc.
1
4
This is intuitive. The same holds for each of the other neighborhoods.
However some people are less embedded into the neighborhoods than others.
The nodes on the edges of different neighborhoods connect the different communities.
1
4
E.g. Scott at the intersection of shitposting, CT, and college tech twitter.
Or Sam at intersection of startup/tech twitter and college tech twitter.
Or at intersection of CT/startup twitter.
Many others....
Pretty cool!
2
6
"Small-world" networks are characterized by the "small-world phenomenon" which you might have heard as the "6 degrees of separation" thing.
That you're at most 6 (or n < 6) hops away from anyone. This seems to mostly confirm this, "small-world" networks have short-path lengths
1
1
This also gives some insight about how information flows through different parts of the network and why things go viral.
Information first spreads within neighborhoods. Connecting nodes then spread this across different/new neighborhoods, which you may not be part of
1
3
"Small-world" networks have two properties we care abt.
L(p) = degree of separation between any two nodes in the network
C(p) = how dense is each local neighborhood.
Certain neighborhoods are significantly more dense than others e.g. ML Twitter vs. General Tech Twitter
3
2
Certain neighborhoods are more withheld than others.
E.g. look at how much overlap there is between crypto / general tech twitter (blue vs. green), and college tt vs. general (purple vs. blue).
In comparison to ML twitter and general tech or crypto twitter (black and others).
1
4
And part of that is just who I follow and who follows me for sure, but still super cool!
If you want to read/watch some good technical material on this stuff
nature.com/articles/30918
youtube.com/watch?v=UX7YQ6
barabasi.com/f/67.pdf?curiu
n/n
7