The outstanding thing about China’s 600 million people is that they are “poor and blank” . . . . On a blank sheet of paper free from any mark, the freshest and most beautiful characters can be written; the freshest and most beautiful pictures can be painted.

– Mao Zedong

Some people love science for its own sake, but most of us value it because it enables good engineering. Few there be who enjoy reading circuit diagrams or materials science treatises, but everyone wants a powerful smartphone. Almost no one wants to wade through the equations of hydraulic physics, but people spend more than half a million dollars every day to cross the Lincoln Tunnel.

For social science, this relationship between science and engineering is nearly reversed. Non-scientists are more likely to share academic social science research findings with their friends than they are to share any other type of scientific research, indicating that many laypeople find social science interesting for its own sake. And on the other side, “social engineering” is an ugly term that scientists disown and that makes most people uncomfortable. Alone among the sciences, social science is prized more for its interesting ideas (science) than for its downstream technologies (engineering).

Princeton University Press’s ongoing Analytical Sociology book series provides reasons to be both excited about social science ideas and uneasy about their application, as most of us are. Damon Centola, of the University of Pennsylvania, has written How Behavior Spreads as the third book in the series. The book opens with a puzzle: Why is it that HIV has spread rapidly around the world but relatively easy behaviors that could prevent HIV infection have not? It is posed as a social scientific question, but it is easy to think of a social engineering analogue, something like: How can powerful institutions manipulate people’s behavior (to minimize the spread of HIV)?

The later chapters of Centola’s book are written to answer just these types of engineering questions, and the book is clear and innovative enough to serve as a primer on the implementation of social control by the powerful. But like all good social science, the ideas are more exciting than the implementation. While traditional behavioral research has either focused on individual psychology or population-level trends, this research occupies a fascinating place in between, examining the implications of the network structures that people form when they interact and connect.

The image below shows the types of networks that the research focuses on.¹ In these graphs, each dot represents a person, and the lines between the people represent social connections. In the “regular” graph, each person knows his neighbor and his neighbors know each other, but no one knows anyone who is spatially distant. This is a simplistic version of what we might imagine the world looked like before the transportation and communication technologies of the last few centuries, when most people never traveled more than 50 miles from where they were born. It is a “large world” because if one person has a virus or letter or idea, it takes a maximally high number of steps to transmit it, neighbor to neighbor, to a distant person.

Watts-Strogatz structures

The “Small-world” graph in the middle is quite similar to the “regular” graph: nearly everyone knows his neighbor and nearly no one knows anyone who is spatially distant. However, there have been a small number of “rewirings,” each maybe representing a person who has neglected his friendship with his neighbor in favor of a pen pal in Russia or an ex-girlfriend in Tahiti. Now that there are a few connections that cross wide spatial divides, the number of “hops” it takes to transmit a virus between any two randomly selected people is greatly reduced. In the “Random” graph, connections show no spatial pattern whatsoever, and the number of steps needed to transmit something anywhere is minimized—it’s the “smallest world.”

These network structures were popularized among academics by Duncan J. Watts and Steven Strogatz in a 1998 Nature paper. Their key insight was that even though the “small world” graph is extremely similar to the “regular” graph, its two or three rewirings were enough to make transmission across the world almost as efficient as the “random” graph. In other words, even if only a tiny percentage of the population is connected to someone far away from them, anyone can get a virus or an idea from anyone else extremely quickly.

The small-world notion is intuitively appealing. It is thrilling to think that even without a cosmopolitan personal network, one is only a few handshakes away from a Congolese peasant or the Sultan of Brunei or Kevin Bacon. On social media platforms like Twitter, it is easy to connect with strangers half a world away, and the social rewiring this creates has been used to explain everything from the success of new products to the Arab Spring. As more people travel or interact with distant strangers online, the world gets “smaller” and in theory it is easier for ideas and behaviors to “go viral” and spread quickly.

Except when it’s not. Centola documents a huge variety of cases in which ideas and behaviors have not spread via a small-world model of jumping across space and social divides. Rather, Centola shows that quite often large-world, old-fashioned, neighbor-to-neighbor transmission without immediate hops across space is both faster and more successful. This violates both intuition and the theory of contagion that used to be accepted.

To explain this anomaly, Centola introduces a distinction between simple and complex contagions. Measles and rumors are textbook examples of simple contagions: A single exposure is sufficient for transmission. They can travel quickly by airplane and spread the fastest in smaller worlds. A complex contagion requires multiple sources of reinforcement in order to be transmitted. Membership in social movements, behaviors related to health, and large investments are complex since one wouldn’t be likely to do them until one had been influenced by several connections rather than just one.

If three neighbors start a new political party, according to this theory, the party’s spread through a large world (like the one on the left of the image above) is straightforward and efficient. A person convinces his neighbor, with the help of another neighbor who is a mutual friend. In a large world, networks are clustered and one’s friends know each other. This clustering and the ubiquity of mutual friends make possible the multiple contacts that are necessary for the transmission of complex contagions. In the smallest possible world (like the one on the right of the image above), a political party will never expand beyond its founding members, because they have no mutual friends and so cannot “team up” to provide multiple sources of reinforcement to attract new recruits. Hence Centola’s central insight: While viruses and simple contagions spread the best in small worlds, new movements and other complex behaviors will spread best in large ones.

This result is especially important since the life-altering behaviors that we care the most about tend to be complex contagions. Centola identifies large-world-style behavioral contagion in contexts as diverse as birth-control practices in Korean villages, trade unionization in Northern Europe, participation in 1964’s “Freedom Summer,” and innovative house construction methods in Kenya.

The results of the book provide a strong reason to be leery of the rise of online social technologies. Every year fewer Americans know their neighbors well, and in general the internet makes our world smaller. This makes it harder for complex contagions to spread through our increasingly weak ties with each other. Centola describes the likely result of this change as

a form of social amnesia. . . .The everyday memory of how people interact and the kinds of gestures or civic-minded behaviors they are expected to display may be transformed. . . .While simple contagions may be conspicuously better spreaders, they typically are not very effective for. . . .transmitting new ideas that will improve the common welfare.

In other words, the metaphor we have of “going viral” is apt: things that spread like viruses are simple, lowest-common-denominator, and require little investment or discipline, and most important they tend to be harmful like viruses. The trend of the world’s social network structure is towards enabling those types of contagions to spread more easily, and to make complex, socially beneficial contagions rarer and more often stopped in their tracks.

Centola’s careful analysis of network structures and the way that behavioral contagions spread may also provide a way to understand more intangible cultural trends. For example, a variety of commentators have asserted that our culture hasn’t generated distinctive innovations since about 1990, in fashion, art, popular music, and culture overall. Complex cultural practices are precisely the types of contagions that Centola claims are more difficult to spread in our increasingly small world. Could it be that the rise of the internet has caused us to live in a permanent cultural stasis or devolution in which every year is a slightly worse version of 1992? The question is too big to answer definitively, but Centola’s ideas would provide a strikingly apt explanation for this if it were true.

Part III of the book is about “social design,” a term more palatable but roughly synonymous with “social engineering.” Centola has been a pioneer in experimental sociology, and has found ways to connect members of online communities to each other to push the overall networks towards resembling either large or small worlds. In these artificially constructed networks, a few random nudges to some carefully chosen “seed” individuals can create cascades of behavior change, for example new diet or exercise practices that spread through the whole network.

The success of these experiments to spread complex behaviors invisibly among crowds of strangers is both impressive and discomfiting. With each advance in social science, some new ingenious method to control other people’s behavior becomes possible. If the science of behavior control experiences enough serious breakthroughs like Centola’s, we can imagine a future in which human liberty is curtailed by powerful groups who “weaponize” social science to bend crowds to their whims.

The desire to use social science for the sake of behavior control is not new. As Theodore Dalrymple has ably pointed out, it is even a plot line in Hamlet. King Claudius summons Rosencrantz and Guildenstern and points out “Hamlet’s transformation” of his “inward man.” He hopes that they “may glean, whether aught, to us unknown, afflicts him thus, That, open’d, lies within our remedy.” He wants to understand Hamlet’s psychology and behavior for the sake of using a “remedy” to control it—it is science for the sake of engineering again.

Hamlet obviously resists both the attempts to understand and to control him. He asks Guildenstern to play on a pipe, and after Guildenstern refuses because he lacks the skill, rebukes him:

Why, look you now, how unworthy a thing you make of me! You would play upon me; you would seem to know my stops; you would pluck out the heart of my mystery; you would sound me from my lowest note to the top of my compass: and there is much music, excellent voice, in this little organ; yet cannot you make it speak. ’Sblood, do you think I am easier to be played on than a pipe? Call me what instrument you will, though you can fret me, you cannot play upon me.

Hamlet has rightly identified the intention of Rosencrantz and Guildenstern—the social scientists of the Danish court—to control him by understanding him. As social scientists in the centuries since have learned more of the “stops” of the human mind, they have gotten closer to a future in which they can play people like instruments. They are, as earnest scientists, trying to pluck out the heart of the mystery of human behavior, but as they try they are also getting closer to being able to control us.

Nor is this only a hypothetical concern for playwrights and paranoiacs. The World Bank and other powerful institutions continue to support the creation and development of governmental “nudge units” that seek to take advantage of the quirks of human psychology to control subject populations. Numerous private consulting companies charge huge fees to help them.

An understanding of social network structures casts new light on recent controversies. Eminent domain takings like the infamous Kelo v. City of New London of 2005 provide a case in point. These seizures push our social world to be smaller by destroying dense networks and large-world communities. The reasoning of judges who have allowed such takings is frequently some version of a public benefit vs. private harm argument: that the public (for example, economic) benefit is great enough to justify the private harm caused by displacing citizens from their private property. With Centola’s research, we can identify a serious public harm—the destruction of socially beneficial large-world networks—in addition to the alleged public benefits and private harms.

Outside of the United States, there are numerous recent examples of policies that have had (mostly pernicious) effects on citizens’ social network structures. In Singapore, a housing law was passed that directly prevents large-world clustered networks from forming by instituting racial quotas in apartment buildings. In China, land seizures have been conducted on a scale that dwarfs New London and other U.S. cases. As in the U.S. context, the victims of such takings abroad tend to be members of politically weak classes.

Advocates of busybody government intervention should remember that the unintended social network consequences of these policies could not have been fully understood when they were implemented. The unintended consequences of the laws we pass this year or next may similarly be unappreciated for decades to come.

Those who love science for its own sake tend to have confidence that it can continue to progress forever without limit. This is an exciting thought if we consider inventing food that can feed all of the hungry and interstellar travel that can carry us to distant galaxies. But for social science, unlimited progress in understanding human behavior would mean the possibility of unlimited control of some by others. It would mean that we could arrive at Mao’s fantasy, in which a central state could write on its citizens like blank paper by influencing and controlling their most important decisions. With recent advances in facial recognition technology, data storage infrastructure, and predictive modeling, Mao’s successors are already beginning to make that fantasy a reality.

Robert Cialdini, the famed scholar of psychology and marketing, once told me that he believed it vitally important for individuals to properly educate themselves about methods of social influence so as to more effectively resist them. For readers seeking to take up the challenge, Centola’s book is a superb place to start.