Going Extreme without Leaders
The United Kingdom has decided to leave the European Union via a referendum last week. This has caused a total shock globally, and many are now seeking who to blame.
However, you may not be able to simply attribute the cause of a system’s (mis)behavior to its individual components. This is one of the key insights produced in the field of complex systems science, an interdisciplinary field of research about systems made of a large number of interacting components. This insight tells us that accusing particular individuals may not be enough to understand and address problems that arise in our society.
In this post, I would like to take this perspective and discuss a few properties our current social system may have. (I should warn you, though, that the discussions developed below are all my naïve personal views, as sociology and political science are not my forte.)
Going Extreme Trending
The recent societal change of the UK toward Brexit was rapid. The idea of leaving the EU was beyond insane only a few years ago. But in a very short period of time, the Brexit movement, combined with the rising nationalism, has gained so much momentum that it has turned the outcome of the last week’s referendum from Remain to Leave.
There are some similar phenomena recently seen in other parts of the globe as well. The US presidential race would be one example. I think most of the readers still remember that Mr. Donald Trump was considered only one of fringe candidates when he announced his candidacy last year. And his unexpectedly successful campaign is not the only example in the US politics. It has also been found recently that the disagreement between the two parties in the US House has grown at an exponentially faster pace over time [1], indicating that the lawmakers are becoming more and more polarized and non-cooperative than ever before at an increasing speed, with no signs of reversal. If I am allowed, I would also like to add another (perhaps controversial) example, which is the distributed self-radicalization of terrorists promoted by information propagation on the Internet and social media.
What I see in common in those, say, “extremization” processes, is that their speed is very fast, and it proceeds in a distributed fashion.
There are numerous examples of the rise of extreme ideologies and regimes in human history, but they were mostly caused by extreme elites in a top-down manner. Compared to historical cases, the recent examples mentioned above are different in that they seem to show a rapid, bottom-up popular extremization. I have been wondering how to understand this. Of course, in some cases popular extremization may be induced deliberately by elites to accomplish their political goals. That would be an easy case where we could simply blame those who initiated it. But I don’t think such a simplistic, causal explanation would be sufficient to explain what have been happening in our society lately, and to develop potential solutions to address them. We may need a more complex-systems-oriented approach.
Coevolution of Social Structure and Opinions
The fast-paced, distributed extremization processes may be understood with a dynamical network model of opinion formation, which is often used in complex systems and networks research [2]. This model consists of individuals and their connections. The individuals may include voters, politicians, media channels, journalists, and any other actors that participate in the social opinion formation process. Each of the individuals has his/her own opinion state, which can change dynamically over time. The connections represent channels of information propagation from source individual to receiver individual. The connections can also change dynamically over time. How those states and connections change depends on specifications of a model.
Here, I consider three underlying mechanisms of how opinion states of individuals and connections among them change over time, as a potential explanation of the distributed extremization processes we are witnessing.
The first mechanism is what researchers call “homophily”, i.e., the tendency of people to connect to others that are similar to themselves. This means that the structure of a social network will be affected by the states of individuals in it. This naturally causes the formation of social “bubbles”, as often reported and discussed at a number of places recently. We tend to hang out with like-minded others. We tend to connect to friends and other information sources on social media that provide us with information that meets our taste. This makes your timeline filled with all similar kinds of information of your favorite, which, while certainly pleasing to us, makes us lose the opportunity to get exposed to different views and perspectives with which we may disagree. What each of us is seeing as the state of the world is, most likely, no more than a local state of a tiny, homogeneous community around you.
The second mechanism is the opposite of the first one, i.e., the tendency of people to get influenced by others around themselves. Called “social contagion” or “social diffusion” in the scientific literature, this mechanism means that the states of individuals will be affected by the structure of the social network they are in. We tend to perceive a “social norm” as the average of opinions exchanged within the social bubble we are in, and our own opinions are naturally dragged toward it, consciously or unconsciously. This leads to the formation of shared mental states among the members of a community, which may be good (e.g., well-established corporate culture) or bad (e.g., suboptimal groupthink in a team), depending on the context.
These two complementary mechanisms (homophily and social contagion) can occur simultaneously, and that can explain some interesting social behaviors. For example, it is known that, under certain conditions, the interplay between these two mechanisms can lead to a fragmentation of society into multiple disconnected communities within each of which opinions are fairly homogeneous but across which opinions differ significantly [3]. Social fragmentation interferes with information exchange at whole society levels and prevents discourses among different communities from happening. Moreover, it always gives people inside each community an illusion that their opinion is the mainstream that well represents the entire society. This might be one of the reasons why the “Remain” supporters in the UK were so shocked by the result of the referendum.
However, combining homophily and social contagion is not enough as an explanation of the distributed extremization that I want to understand. This is because they do not account for possible dynamics within each bubble that may drive their opinion to extreme. We need something else.
Pay Attention to… Whom?
The third underlying mechanism I would like to bring in as a missing piece, is people’s inherent nature to pay attention to things that are distant from the average. When we see something that stands up among many other similar average-looking options, we simply can’t help but pay attention to it. This is quite a natural behavior; average options wouldn’t give us any surprises, but things that are distant from average will certainly give us some surprise (= information). Such a strategy of information gathering must have been beneficial for animal survival and thus evolutionarily favored. I am not sure if there is an established technical term for this mechanism since I am not so familiar with psychology or cognitive science. Here, I tentatively call it “selective attention” for now.
When an action of selective attention happens, it modifies the structure of the social network a little. The information source that exhibited an opinion that stood out among others will gain a little more recipients than before. This modification of social structure can occur regardless of whether either the source or the recipient had such an intention, or whether the opinion was liked or hated by the recipient.
One thing that is quite important to note here is that, if the social network structure changes a bit inside a bubble, the perceived average opinion, or social norm, within that bubble also changes a bit, too. This is highly nontrivial, but mathematically true. Even if no one’s opinion has changed yet, a modified network structure changes the weight by which each individual’s opinion contributes to the perceived average, causing a slight drift of the social norm in the bubble [4].
If this selective attention mechanism is combined with the other two, what will happen? Society can self-organize into several disconnected bubbles, in which, among many similar-looking, homogenized opinions, a little more extreme opinion that stands out naturally attracts more attention and causes a little drift of social norm, which then spreads among the members of the community via social contagion. Repeat. Then we will find that we are reaching a very bizarre scenario: Even if everyone tries to assimilate with their social neighbors and if no one has any intent to steer the society to any desired direction, it is possible that society can self-organize into disconnected, extremely opinionated bubbles.
Who is to blame? Probably no one. In this scenario, everyone tries to blend in and conform to their respective social norm. But the dynamics of society as a dynamical network could make such an emergence of extremism without leaders possible.
Think Average
I am not claiming that the model and the conclusion derived above are empirical truths by any means. Apparently, many cases of extremism were/are led by a small number of elite leaders and/or by exogenous political or economic factors. My only point is that, by considering other potential mechanisms of self-organizing distributed extremization, we may be able to free ourselves from a simplistic view of attributing causes to individuals. This may help us to take more concrete responses and actions at our own individual levels, instead of only blaming those accused individuals.
What can we do specifically to prevent our society from going extreme? I believe the model discussed above could give us some insight into this question. Simply put, we can try the opposite of each of the three mechanisms. But this is easier said than done.
Doing the opposite of homophily means you try to keep connections with the information sources with which you disagree, so that you won’t get stuck in a social bubble of your own. But this actually causes a dilemma because, if you keep listening to what they say but they stop listening to what you say, then their social influence becomes greater than yours, which you probably wouldn’t want to see. Unless the society as a whole adopts a non-homophilic behavioral rule, preventing homophily is not practically doable due to this dilemma.
Having said that, doing the opposites of social contagion and selective attention can be practiced at individual levels without causing such a social dilemma. Namely, you try not to easily absorb your social neighbors’ opinions (to suppress social contagion) and not to pay much attention to those who express outlier opinions (to suppress selective attention). You can adopt these behavioral guidelines all by yourself alone today, and you will immediately contribute to the diversity and moderation of opinions in our society. It seems to me that these are among of the basic literacy that everyone who lives in a highly connected modern society should understand, train, and adopt. They are probably beyond what is typically expressed as “critical thinking” in today’s educational systems, because they have implications of individual actions for networked effects at societal levels.
As a final remark, I would like to show you a commencement speech that has been the most popular video in the YouTube channel of my employer, Binghamton University (trust me, I am not told to do this by them). Here it is:
“Average is the new exceptional”
https://www.youtube.com/watch?v=ULRosL7AOpk
An “average” student speaker Anthony Corvino gave this funny speech at Binghamton’s Fall 2009 Commencement. I like it a lot by its own right, but given the context discussed above, I think this speech also carries a particularly important message. Perhaps it is time we nurtured a culture in which people respect, and value more, moderate opinions and manners that may not necessarily stand out as “exceptional” in a traditional sense.
Further Readings
[1] Andris, C., Lee, D., Hamilton, M. J., Martino, M., Gunning, C. E., & Selden, J. A. (2015). The rise of partisanship and super-cooperators in the US House of Representatives. PLOS ONE, 10(4), e0123507. http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0123507
[2] Sayama, H. (2015). Introduction to the Modeling and Analysis of Complex Systems. Open SUNY Textbooks, Milne Library, Chapter 16. http://textbooks.opensuny.org/introduction-to-the-modeling-and-analysis-of-complex-systems/
[3] Kozma, B., & Barrat, A. (2008). Consensus formation on adaptive networks. Physical Review E, 77(1), 016102. http://journals.aps.org/pre/abstract/10.1103/PhysRevE.77.016102
[4] Sayama, H., & Sinatra, R. (2015). Social diffusion and global drift on networks. Physical Review E, 91(3), 032809. http://journals.aps.org/pre/abstract/10.1103/PhysRevE.91.032809
