You might wonder: what does mind viral immunity have to do with network graphs?
Well, in fact, they are closely linked. So closely, that graphs can be used to measure the level of any system's mind viral immunity, as well as its capacity to contaminate.
If you apply this approach to text network analysis, you will be able to estimate contagious potential of any text — or, alternatively, see how "immune" a certain discourse is to external influence.
We will demonstrate how it works below using the social network analogy.
The graph above is a network of the Twitter users who were tweeting around the topic of Neuralink.
The more dispersed this network is, the harder it is for it to synchronize, the longer it will take for any message to pass through. Moreover, there are several elements disconnected from the rest. Therefore, this network is almost too immune: it's very hard to not only contaminate it, but to also unify it for a collective (informational) action.
Mind-viral immunity score: very high.
Let us now consider a fictional example: suppose this social network grew and evolved over time and became very interconnected.
It would look something like this:
It's much easier for information to travel through this more interconnected network structure, so the network is much less immune and can be contaminated faster (Kuperman & Abramson 2001, Zhou et al 2007). If @elonmusk was to tweet something, the message would quickly reach all the parts of the network. The advantage here is that information spreads fast and it is possible to recruit people for a collective action. You want to have a social network like this if you were to start a revolution. It easy propagates contagion and is also contagious itself.
Mind-viral immunity score: low.
We want to be able to adjust our immunity — and our contagion potential — depending on our circumstances and objectives.
Sometimes we want to be part of a more contagious environment and sometimes — less. That's why it's important to be able to know how contagious / immune a network is, so you know where to belong and how to adjust, if needed.
Using InfraNodus you can transfer this approach in the context of discourse analysis and estimate the level of mind viral immunity for a discourse.
Step 1: Add any text to InfraNodus (your own, an article, news, search results, tweets on a certain topic)
Step 2: You will see a graph visualized where the nodes are the words and their co-occurrences are the connections between them (Paranyushkin, 2019):
Step 3: InfraNodus will estimate several parameters to give a score to the resulting discourse network structure:
- graph's modularity (how pronounced is the discourse's community structure)
- distribution of influence over topics (how well are the influential terms distributed over the topics)
- percentage of nodes in the most present topic (too high amount means bias)
If the graph's modularity is high and influence is evenly distributed (i.e. its entropy is high) then the Network Structure is marked as Diverse or Dispersed and its Mind Viral Immunity score is High or Very High. This means that the way this discourse evolves over time is quite diverse. This is our case above. The discourse above is represented by a diverse network, which makes it slower to propagate and sync and — at the same time — more resilient to external contamination (Pastor-Satorras & Vespignani 2000, House & Keeling 2009)
Alternatively, if the modularity is low (the topics are too well interconnected to the point that they are dissolved among the most influential terms) or if the focus is biased towards one topic, the discourse is then very interconnected and — thus — can easily and quickly propagate, it also has a high contagion potential and low mind viral immunity (analogy: highly ideological short texts will have this score low — Paranyushkin 2018).
Step 4: You can then estimate what you want to do with this discourse.
For example, if your intention was to present a point of view in a very focused way a "Very High" score above may indicate that you need to make this discourse (text) more interconnected.
Alternatively, if the score was low and you were analyzing the news discourse that would mean that the world badly needs alternative takes on the current situation, so you could create relevant content to introduce more diversity in the discourse.
Mind viral immunity, even on the most general level that Elon Musk was talking about on Joe Rogan's podcast, is an important concept to consider. In the future where the Neuralinks — small electronic implants — become omnipresent and are used to communicate, it will be very hard for people to filter the knowledge they receive and exchange. Filter bubbles and echo chambers, fake news and deep fakes will proliferate. In a densely interconnected network it will be very easy to fall prey to a fad or for an ideology, so the possibility for totalitarian narratives to rise up increases.
One of the possible ways to resist this is to ensure a certain dynamic level of diversity in a system. Diversity makes it resilient, however, the distinct parts should themselves be dynamic: both in their interaction with other parts and within themselves.
An open system will both include and let go; it will develop, but also decline. It will accept a cyclical nature of things, but, still, with an evolutionary force within. Without any desire to overtake the rest, but, rather, co-exist, in a way that is not too connected and not too disconnected. A comfortable level of mutual co-immunity.
Networks provide a very useful way of representing this dynamics in a way that can lead to tangible insights as to what should be done to ensure things evolve in this direction.
Take a look at this dynamic visualization of the news of the day taken from four different sources: The Guardian, The NY Times, The Financial Times and The Washington Post. You will see that while they cover on a lot of similar subjects, The Washington Post, towards the end, introduces diversity into the day's news discourse, bringing the topic of Sports into the context dominated by Coronavirus and Trump. Its present in the analysis makes the Mind Viral score High. It if it was there, the score would be lower as too much concentration would be on a handful of subjects.
If you are interested in a deeper discussion of this subject, check out a more comprehensive discussion of Mind Viral Immunity on Nodus Labs' website.
House, T., & Keeling, M. J. (2009). Household structure and infectious disease transmission. Epidemiology and Infection, 137(5), 654-661. Cambridge University Press.
Kuperman, M., & Abramson, G. (2001). Small World Effect in an Epidemiological Model. Physical Review Letters, 86(13), 2909-2912. doi:10.1103/PhysRevLett.86.2909
Paranyushkin, D (2018). Measuring Discourse Bias using Network Analysis. Towards Data Science (link).
Paranyushkin, D (2019). InfraNodus: Generate Insight Using Text Network Analysis. The World Wide Web Conference 2019, 3584–3589, doi:10.1145/3308558.3314123
Pastor-Satorras, R., & Vespignani, A. (2000). Epidemic spreading in scale-free networks, 13. doi: 10.1103/PhysRevLett.86.3200
Zhou, J., Liu, Z., & Li, B. (2007). Influence of network structure on rumor propagation. Physics Letters A, 368(6), 458-463. doi:10.1016/j.physleta.2007.01.094
Please sign in to leave a comment.