Network Dynamics framework for Social Network problems

Note: You may notice that this post is dated older than when this website was created. That is because, this is a cross post. I had originally posted this in my old blog here. I am cross-posting it here in the spirit of collecting everything I have written in one place.

Social Networks are all the rage today (in many senses of that phrase). What we once thought would democratise content creation, is now breaking democratic norms. Given the scale of the problems involved, it is becoming increasingly difficult to think of them. Here I have made a humble attempt to do so and make the case for the framework I have been using. I am only beginning to think deeply about this issue so I might be wrong about this and change my mind later but I believe there is some utility to sharing it anyways - even a wrong clock shows the right time twice a day! So here goes nothing…

Network theorists classify network dynamics into 3 categories - dynamics on networks, dynamics of networks and hybrid dynamics [1]. I believe we can use a framework with a similar classification to categorize problems relating to social networks and to keep in mind all the dynamics involved before creating solutions.

Terminology

Since I have begun by throwing around some jargon, I now feel obliged to explain them before proceeding. Feel free to skip it, if you are already familiar. Networks are representations of systems in which multiple agents interact with each other. Each agent in a network is called a node and their interactions are called edges. These nodes and edges can have certain properties attached to them. Let us consider the case of a social network. In this case, each node would be a person and if these two persons interact with each other in some way (for example ‘Comment’ on a post) they would be connected by edges. Note that if two persons never interacted with each other then there would be no edge connecting them. Property of a user could be ’number of posts’, ’number of likes’, ‘Joining date on the platform’ etc.. Property of an interaction could be ‘Retweeted’, ‘Commented’, ‘Liked’ etc..

Network dynamics is the study of how a network changes with time. As aforementioned, there are three distinct ways in which network dynamics can play out:

  1. Dynamics on networks: This is when the number of nodes or edges doesn’t change with time but only their properties change. For instance, imagine studying a network of students in a classroom. Typically new students don’t join the class or leave it in the middle of a school year. So the number of students (nodes in the network) remains the same. Also if the class is small enough, everyone is at least acquainted with everyone else. So everyone has interacted with each other and so the number of edges is at a maximum and can’t change either. Now, in such a network, the dynamics would only be changes in properties of the students themselves such as grades or in properties of their interactions such as friendships, rivalries etc. over the school year. This is an example of dynamics on a network.
  2. Dynamics of networks: This is when the number of agents and interactions itself changes over time. For instance, when Facebook started out it connected students in difference colleges and continued to grow. Here we have multiple classrooms now starting to get connected together hence creating a possibility for new students (nodes) entering or old students exiting the network. Also there are now students in the network who possibly don’t know each other at all creating opportunities for new interactions (edges) in the networks. The patterns of change in the entry/exit of students and the creation/removal of their interactions here, constitutes dynamics of the network.
  3. Hybrid dynamics: This is when both the dynamics occur simultaneously. Imagine the splitting up of a political party into two factions. At the beginning, both the factions would have had the same ideology. But over time, the ideology of some of the members started changing and this change made others in the party rethink their’s (dynamics on the network), which eventually lead to its split up (dynamics of networks). Similarly, in a merger of companies with two different cultures, the emergence of a unified culture over time is also an example of hybrid dynamics playing out on a network.

The framework

With the terminology now clear, here is what the Network Dynamics framework dictates - When we come across a problem in a social network, it is first important to recognize all the dynamics involved. If we focus only on some but not others, in due time we will be forced to confront them too.

Let us take the example of Information disorder - the problem of spread of mis/dis/malinformation in an online social network. On the face of it, there is an existing network of people and the property of them acquiring some information alone is changing over time. So, it is tempting to categorize this as an example of dynamics on networks. But one could argue that since bots are created (new nodes in the network) to spread the information, there is dynamics of networks also here. This argument could be further strengthened by the point that mis/dis/malinformation often creates heated arguments wherein people who typically wouldn’t interact, will now end up interacting which will cause new edges in the network too. So really, the Information disorder problem starts out as dynamics on networks and then causes dynamics of networks. On the whole, there is Hybrid dynamics involved here.

But say we indulge our initial temptation to only consider the dynamics on the network which causes Information disorder. Then, our only focus will be to change the way people acquire information. For instance, we would recommend that critical thinking is the only solution. But as Danah Boyd says [2], this can backfire. When there is so much information flowing in, some information just needs to be taken for granted if it comes from trusted sources. When people don’t do so and keep resorting to extreme critical thinking, the issue simply persists. But why is there so much information flowing in the first place? Well, that is because there are so many new nodes and edges created by dynamics of networks and we need solutions for that one. And there it is - we started out by just looking at one dynamics and then end up being forced to confront the others as well.

The same will happen if we start out looking at only dynamics of the network too. I believe the proposal to break up social media platforms like Facebook [3], stems from looking at dynamics of the network alone. First off, Information Economics suggests that social media platforms will simply tend to reach critical mass and so even from the standpoint of dynamics of the network it doesn’t make sense - even if we break it up, one of the broken pieces will simply reach critical mass again and become the new Facebook [4]. Secondly, it simply doesn’t address Information disorder at all. Even smaller groups can share mis/dis/malinformation and then by creating engagement through inflammatory posts become big. At the end of the day, it is caused by dynamics on networks. There it is again - start with only some of the dynamics in mind and we will be confronted with the others very soon.

Just a problems framework

I hope the framework makes sense. But as it stands now, it only gives a perspective on the problem. It is still not talking about how to go about solutioning for each of the three dynamics involved. One obvious perspective on the solutions that it brings out is that, it is very easy to have unintended consequences given the potential for hybrid dynamics to play out. But that is as far as it goes for now, in terms of solutions.

This is where more work is needed - is it possible to come up with some general recommendations for problems of each of the three dynamics? If one can answer that question, then once all the dynamics are recognized in the problem, policies can take multi pronged approaches to solve for each of the dynamics and a holistic proposal can be built. I am going to continue thinking about this and would love to hear your thoughts on the same.

References

[1] Sayama, Hiroki. Introduction to the modeling and analysis of complex systems. Open SUNY Textbooks, 2015.

[2] “danah boyd: How Critical Thinking and Media Literacy Efforts Are ‘Backfiring’ Today | EdSurge News.” [Online]. Available: https://www.edsurge.com/news/2018-03-07-danah-boyd-how-critical-thinking-and-media-literacy-efforts-are-backfiring-today. [Accessed: 21-Dec-2020].

[3] “Opinion | It’s Time to Break Up Facebook - The New York Times.” [Online]. Available: https://www.nytimes.com/2019/05/09/opinion/sunday/chris-hughes-facebook-zuckerberg.html. [Accessed: 21-Dec-2020].

[4] “Breaking Up Facebook Won’t Fix Social Media.” [Online]. Available: https://hbr.org/2020/09/breaking-up-facebook-wont-fix-social-media. [Accessed: 21-Dec-2020].