Nicholas Christakis: The Science of Social Connections (HeadCon '13 Part V)

[Video designed for full screen viewing. Click icon  in video image to expand.]

Nicholas Christakis: The Science of Social Connections 

If you think about it, humans are extremely unusual as a species in that we form long-term, non-reproductive unions to other members of our species; namely, we have friends. Why do we do this? Why do we have friends? It's not hard to construct an argument as to why we would have sex with other people, but it's rather more difficult to construct an argument as to why we would befriend other people. Yet we, and very few other species, do this thing. So I'd like to problematize that; I'd like to problematize friendship first. 

Nicholas Christakis is a Physician and Social Scientist; Director, The Human Nature Lab, Yale University; Coauthor, Connected: The Surprising Power Of Our Social Networks And How They Shape Our Lives.

The part of human nature that I'd like to talk about today is that part of our human nature that is relevant to our interactions with others. There's been a phenomenal amount of work taking place in the last ten years, certainly, and even in the last year or two that seeks to understand how we interact with each other and how we assemble ourselves into social networks.

If you think about it, humans are extremely unusual as a species in that we form long-term, non-reproductive unions to other members of our species; namely, we have friends. Why do we do this? Why do we have friends? It's not hard to construct an argument as to why we would have sex with other people, but it's rather more difficult to construct an argument as to why we would befriend other people. Yet we, and very few other species, do this thing. So I'd like to problematize that; I'd like to problematize friendship first.

Second, not only do we have friends but we prefer the company of people we resemble. There's an enormous amount of literature on in-group bias and on why this might be the case. A lot of this literature, to my eye, takes the form of what I would regard to be a tautological explanation. Why do we prefer the company of people we resemble? Because we're more comfortable when we are with people we resemble. Why are we more comfortable when we're hanging out with people we resemble? Because they resemble us. And I'd actually like to try to find a deeper explanation for why we befriend other individuals; why we assemble ourselves into networks with what turn out to be very fundamental, reproducible topologies (structures); and why we prefer the company of people we resemble.

And, in fact, the ubiquity and necessity of social interactions carries with it a suite of other phenomena, like cooperation, which is very deeply and fundamentally important; sensing (the ability to see what's happening in others); communication; social learning; epidemics; violence – all of these phenomena arise not so much within individuals, but rather at the interstices between individuals. They're not so much nodal phenomena—having to do with the nodes on the networks—but edge phenomena—phenomena that have to do with the connections between the individuals.

In fact, I'd like to think that the focus on networks calls into question some very old ideas about human nature, and about what the state of nature really is for human beings. Joe Henrich, in an interview he did for Edge a couple of years ago, had a very nice, pithy summary of this. He asks why do we see market economies as all about competition for advantage? Actually, you can just rethink the existence of market economies as all about cooperation. Why do we have to see them as being competitive rather than as cooperative enterprises?

We can shift our perspective on lots of things when we think about people as being nodes on a graph, as being connected to other people. And this shift in focus might, in fact, prompt us to begin to think about —not the individuals themselves‑but the ties between them. This calls to mind an analogy, which I don't know if some of you may already know, of streets in the United States and in European countries. So, streets have names in our country, and the houses on the streets are numbered numerically and linearly as you move along the street. And the blocks between the streets don't have names or numbers and are seen as the things that are between the streets, and we don't pay much attention to them. But if you go to Japan, it's the blocks that are numbered. The blocks have names and the houses on the blocks are numbered in the order in which they were built, not numerically or linearly in any kind of systematic way. If you ask the Japanese, "What's going on with the streets?" they say, "The streets are the spaces between the blocks." They don't pay attention to those.

We can even begin to think about human beings in this fashion. We're so interested in understanding human beings that we lose sight of the connections between them. And just like we can efface the individual, to some extent—and I don't have a strong argument that we should do this, but I have what I would regard to be a weak argument why it's beneficial or useful as a heuristic to do this—just like we can begin to efface individuals by thinking about the selfish genes within them, we can also begin to efface individuals by thinking about the connections outside them.

So, the question I'm asking myself lately is: What would a social science of connections, rather than a social science of individuals, look like? What would it mean to take connections as the focus of inquiry and to think about the individuals as the spaces between the connections who are not so important? And then we begin to think about all the dyadic interactions between individuals, which are themselves natural phenomena, just like we are. I'm an object of the natural world, but so are my connections between me and all the other people, so are those connections objects of the natural world which warrant an explanation and a kind of deep and profound—in my judgment—study.

In fact, this would have a variety of conceptual and methodological problems. And some people would say that this is a really horrible perspective, that it obliterates our individuality, that it's dehumanizing, and so forth. But I would retort to that by saying, What makes us think that the ties between us are any less important or worthy of attention than the individuals themselves?

One of the things that we've been doing is asking ourselves what is the reason that we form these ties? What's the function of these edges and these connections between us? One of the things that's very interesting to us is that these edges between individuals, these networks that we form, have properties that are not reducible to the individuals. They offer us a kind of an understanding of emergence, a new kind of emergent phenomena. And these properties, while they are properties of groups, actually, as it turns out, have implications for individuals.

We're so interested in understanding human beings that we lose sight of the connections between them.

Let me give you an example. This is very visual and, given this format, I'm not supposed to use visuals, but I'm going to cheat and use one slide in a moment. Let's say you had 1,000 people, and, on average, they each have five connections, so you have 5,000 ties between them. Mathematically, you could construct a number of ways in which you could organize these networks. You could have a random network where people are jumbled together; you could have a big ring network; you could have a kind of “scale-free” network; you could have the kind of network that we humans actually make (which has a variety of properties). It turns out that if you were designing the network from mathematical principles so that the network would be the most resistant to pathogens taking root within it; so, you say, "I want to organize these people in such a fashion that this group, when so organized, resists epidemics;" whereas, if they'd been organized some other way, these same people who otherwise were identical—had the same immune systems, the same biology—this group no longer resisted epidemics so well. If you wanted to give the group the epidemic resistance property, the way you would organize the people is to give them a property in network science known as degree assortativity. You would make popular people befriend popular people and unpopular people befriend unpopular people. You could give them this property, it would make the network as a whole resistant to germs being able to make inroads.

And I can cultivate this intuition by asking you to think about the airport network in this country. The airport network is degree disassortative. Chicago is connected to lots of small airports but, in the small airports, you can't fly from one to the other; they are disconnected from each other. Whereas people don't have that property. Popular people befriend popular people, and unpopular befriend unpopular. Now, think about which of those two networks, if you were a bioterrorist and you wanted to seed a germ in, which network would the germ spread more rapidly? In the airport network, right? If you start any random node, like an isolated small town, it will go to Chicago, and, in the next hop, it will reach the whole nation. But if you had the hubs and the spokes or the peripheral airports connected to each other, it would be relatively more impervious to a pathogen spreading.

I don't think it's a coincidence that of all the kinds of ways human beings could organize themselves into networks, that's what we do. We evince degree assortativity, and I don't think it's a coincidence that we do that. We assemble ourselves into groups, the group now has this property, this germ- resistance property, which is a property of the group, but which, as it turns out, also benefits and affects us. Now, being a member of that group, we are less likely to acquire pathogens.

And this sets the stage for a set of ideas that we and others have been exploring that shed light on multi-level selection and other kinds of contentious ideas in the biological and the social sciences. And we have a number of fellow travelers on this road—László Barabási, Dirk Helbing, Tooby and Cosmides, Frans de Waal, Nowak, Rand, Santos—people working on these related areas of interactions among animals and people, and what this means. In fact, David Rand and Josh Green and Martin Nowak just had a nice paper this past year — I was asked to highlight some papers—looking at whether you can use time to response as a kind of heuristic for understanding are people intuitive cooperators and rationally selfish, or do they exercise rational self-control over a kind of instinctive greed? The data they presented in that paper, to my eyes, was quite compelling—that we are intuitively wired to cooperate.

James and I published a paper last year as well, also in Nature, where we had the following idea: We said, well, what we would love to do is, if the claim is that there's something deep and fundamental about human social networks and the structure of networks, we would love to be able to go back 10,000 years to the Pleistocene and look at what kind of networks did humans assemble themselves into, before we invented agriculture, and cities, and communication, and so forth?

We did the next best thing to that, which is to map the social networks of the Hadza hunter-gatherers. There's only about 1,000 of them left; only about 500 of them still live in the traditional way. They are a natural fertility population; they have no material possessions to speak of; they sleep under the stars. And when we map their social networks, their networks look just like ours. So, despite all of modern technology, telecommunication, the Internet and everything else, the structural features of their networks are indistinguishable from the structural features of our networks, suggesting to my eye, again, that there is something very fundamental, not just about the structure of our bodies and our minds, but also about the structure of our societies.

This is some of the work that's been going on in a number of fronts the last few years, trying to understand the social interactions, social networks, and the kind of constituent elements of that – cooperation and the like. But then that leads to what I like to call the so-what question. So what if we can understand the structure and function of networks? What can we do with this knowledge, not necessarily to make the world better, but actually to intervene in the world in some way? And if you think about it, that's also one of the tests of science. I mean, as a scientist, can you actually understand the natural world well enough that you can actually seize control of the natural world in some way and make it obey certain fundamental rules?

I'm going to close with some summaries of a few experiments that have taken place over the last couple of years, and then a bigger idea as the final point. Let me just summarize a few pieces of work that are going on in my field that are very cool at the moment. There are two broad categories of work: One category of work is, can we manipulate the structure, the topology, of the network? Can we take control of the nature of the ties between people and drive the network to desired states? The second is, can we manipulate, not the connection, but the contagion within the network? Given the structure of the network, how can we seed the network? How can we introduce information strategically within locations that make the group behave in desirable ways that we specify? Can we show that we've mastered and understood this world well enough that we can actually intervene in it?

One experiment that was done by a former postdoc of mine [Damon Centola], that was published a couple of years ago now, is this. And I have to show you this image. So, this is an image of experimentally constructed networks. There are two networks in this image. There's just no way you could describe these two networks. Both of these networks have 128 people in them, and in both of these networks each person is connected to exactly six other people? So, if you talk to the human beings in these networks, and ask them, "How many friends do you have? " and they say, "I have six friends. " And every one of them in both of these worlds would say, "I have six friends." They cannot tell the difference between the two worlds which they inhabit?

Now, suppose I'm going to infect the person—the yellow dot that's up here—with a germ. In which of these two worlds do you think the germ would spread more rapidly and more completely throughout the network? From the point of view of the individual, there's no way of telling what world they're in, but from the point of view of us, with this God's eye view, we should have an intuition in which of these two worlds is the germ more likely to spread? And the answer is the network on the left. This random assembly means that, ping, ping, ping, in the next step, the germ will spread from the yellow dot to the six red dots, and then from there to the others, and you'll flush through the system, you'll get a blooming of the information spreading or the germ spreading or whatever. And these are things that spread by so-called simple contagion.

Now, I'm going to ask you something different and more difficult. Imagine now what is to spread within the network is not germs or information but, a behavior, for example, smoking cessation or cooperation. Something more complex. It turns out that the world on the right is the world that is more conducive to the spread of such phenomena. So the topology of the network, which can be seen from above, is what's relevant to whether or not these group-level properties can emerge and be sustained. So this was an experiment that was done to show that.

We did an experiment in our lab where we recruited over 2,000 people online, and we brought them into these virtual worlds, and the subjects played a public goods game with people near them, a kind of cooperative game with those around them who they were randomly assigned. Then, we controlled in that world whether or not people could rewire their networks and the amount that they could rewire them, by which we meant not only can you, if you defect from me, can I reciprocate by defecting, or, if you cooperate, I can reciprocate by cooperating, but we gave me another tool, which is that I could cut the ties or form ties to people. So I could form ties to cooperators and cut ties to defectors. And then we manipulated the viscosity with which that could be done.

What we found was that actually we could control the amount of cooperation that emerged in this group of people by specifying the rules of interaction. If we allowed people to rewire their ties just the right amount, then cooperation in the group would appear above and beyond and independent of the individuals themselves and their own tendencies. So we can elicit from the group a property, namely, cooperation, by controlling the nature of interactions. Second experiment.

A number of other experiments have been done with contagion phenomena. So, given a structure of human interactions in an African village, in a trading floor on Wall Street, in schools in the United States, whatever the setting is, can you strategically introduce information in such a fashion that you can get people to behave in particular ways? There was just a paper published by Matt Jackson and Esther Duflo a couple of weeks ago in Science looking at microfinance. So if you want to get the adoption of microfinance in a setting in Indian villages, who do you target so that if you get them to use the microfinance you get the most spillover and the most rapid diffusion of innovation?

If we could find ways of identifying central people using big data or other techniques, and monitor them passively or actively, when we observe a spike in central people it means that an epidemic's about to strike the population. 

Another nice paper that was done by my colleague, James Fowler – and all of the work that I'm describing to you, virtually all of it, has been done jointly with James—is the following: James did a beautiful paper as well last year in Nature where they randomly assigned 61 million people online to a voting intervention and were able to show that actually showing people a very seemingly trivial stimulus drove, not only the individuals themselves to be more likely to vote, but their friends to be more likely to vote, and their friends' friends to be more likely to vote. So he showed a spread of civic-mindedness to two degrees of separation within this massive experiment done with 61 million people. In fact, it's estimated that an extra 300,000 people turned out to vote on that election because of James's experiment. Actually our democracy was improved because of the scientists actually doing their work in that particular occasion.

There's been some other nice work on product adoption using experiments online: how can we get people to adopt products? And we're in the field right now doing some experiments where we've mapped the networks of 32 highland villages in Honduras, and we're trying to see, if we can only reach 5 percent of the people, which 5 percent should we reach so that we get the whole village to change its mind about clean water and nutrition outcomes? And we're randomly assigning the villages to different targeting algorithms. In some villages, we pick 5 percent of the people at random; in other villages we pick them according to one targeting algorithm; and still another according to another targeting algorithm, and we have very promising results from this study.

There's also a sense in which you can now use networks—and there's been some nice work done in the last year or so, summarizing my field—wherein now, instead of introducing information into the system, you can think about networks as kinds of sensors—extracting information from the system. So, for example, if you think about it, just a moment ago, we cultivated the intuition that if you target information to particular individuals, they're going to be more able to spread whatever it is that's happening in the network.

Let me ask you to think about this, since I can't use slides. Imagine a network. There are ties and there are little nodes between them. Most of you probably have an image that, in the middle, there's a kind of jumble, like Christmas tree lights. When you open them up after a year, there's a thick knot in the middle, and there are these little tendrils that spread out to edges, that's what a network kind of looks like. Imagine that I can ask you, "You can be a person in the middle of that, and have four friends; or you could be a person on the edge of that and have four friends. Now a deadly germ is spreading through the network. Who would you rather be? The person in the middle or the person on the edge?" The person on the edge. You have the intuition that the person in the middle is going to be on more paths through the system – and you can formalize this mathematically – and is going to be more likely to get whatever's spreading through the system. This very simple idea was an idea that we exploited by recognizing that if we could identify who were the central people in networks, and passively monitor them, we would have an early warning system for epidemics. So, the epidemic curve is a classic S-shaped curve that goes up like this. That S-shaped curve should be shifted to the left in central individuals compared to random individuals within the system.

So, if we could find ways of identifying central people—using big data or other techniques—and monitor them passively or actively, when we observe a spike in central people, it means that an epidemic is about to strike the population. This can also be done with economic information or any kind of information that spreads through the system. We were able to show that this works with an outbreak of H1N1 flu a couple of years ago now, and in the last year we also showed that it works on Twitter. James and I know nine days before anyone else what's going to be popular on Twitter, because we see it spiking in the individuals that are at particular topological locations within the network.

To sum up, this is new work that has been taking place over the last year or two in my field, which is network studies and the study of social psychology relevant to interactions and in sociology (not all of sociology or all of psychology, just my little niche where I sit), and the biology of these types of things, has a number of features. First, this work is increasingly experimental in nature; so, more and more people are doing experiments. This move to experimentation is a kind of rediscovery of a tradition of experimentation in the social sciences. We always did experiments, but beginning in the 1950s, we became besotted with regression models. Psychology is a bit of an exception because they consistently have done experiments. But we're moving back to field experiments in broader swaths of the social sciences, and this is being abetted in part by the development of the Internet and online experimentation. So the big data revolution intersects with the experimental revolution by making it easier for us to do experiments.

This new work reflects four things: First, it’s experimental. Second, it's exploiting online and Internet technology. Third, there is (to my eye at least) an increasing desire to try to find things that are deep and fundamental about our humanity. The best social science now that is being done seeks to go to a deeper, more fundamental level to try to explain human behavior, at least when it comes to human interactions. And, fourth, this work is involving interventions. If you want to construct an almost Popperian sort of theory of science, the ability to actually...: we observe the system; we have a hypothesis about the system; we do experiments about the system and conclude things; and now we actually manipulate the system (we introduce genes, we excise the genes, we do experiments in particular ways): this shows a level of control or understanding that's very commendable.

Collective behavior has always captivated people's interest, but, in the last couple of years, we've been making phenomenal progress in understanding what I would regard to be – for me at least – the key aspect of our human nature, which is our interactions with others.

BROCKMAN: You mentioned point three. Could you repeat that?

CHRISTAKIS: Something that transcends individuals. Something that's very deep and fundamental, but that transcends individuals.

BROCKMAN: I think you mentioned humanism?

CHRISTAKIS: I didn't mention humanism. No, but you're asking ...

BROCKMAN: You were imputing some kind of goodness?

CHRISTAKIS: No, I'm avoiding that because those were my marching orders from you, John.

BROCKMAN: There is a sense in all the discourse about networks and big data that it means good. But as Steven Pinker pointed out, the Internet hasn't changed much in terms of human nature.

CHRISTAKIS: Yes, I think I know what you're talking about. Any technology—atomic power, guns—can be deployed for good or for evil. So, I've been highlighting or imagining some ways in which a better understanding of social interactions can be exploited for good. But it can also clearly be exploited for bad. Now, this bad could be getting people to buy products they don't need; it could be whipping up political fanaticism. Actually, if you understand networks, you can be much more effective at fostering Nazism. Actually, there's a way in which you can think of extreme political ideology and how it takes root in populations, and how you would go about structuring populations precisely to reinforce these kind of extreme ideologies. So there are all kinds of bad things that you can use the same technology for, and I am not unmindful of that. But I mean, the things that we're trying to do I would think, we're trying to increase cooperation, and make people healthier, and increase economic development in the developing world, and everything else that Sendhil and everyone else here is trying to do.

SANTOS: I can't help but ask the psychologist question, which is a chicken and egg question, which I'll illustrate with chickens. So, imagine you ran your network analysis on chickens. I don't know what chicken networks look like ...

CHRISTAKIS: Someone's done that, by the way, but go on.

SANTOS: They don't look like humans, right?

CHRISTAKIS: No, they don't. But elephants do.

SANTOS: The primate stuff we're getting out of Cayo Santiago suggests but other animals form networks too, but the question is that why is that, right? And so, you started by talking about this fact that humans might have networks that are unique or unique to more closely related primates or whatever, but then why at the psychological level could that be? Is there something about human cognition or human cognitive mechanisms that allows us to form those networks, and not other species? And, if so, then it seems to me that the individual, at least what's going on in the individual's head, shapes this

CHRISTAKIS: I think it's fascinating. Leaving aside the eusocial insects and clonal species, where the interactions between the individuals are necessarily amongst kin, we're talking about non-kin relations, so we've got primates, including us, elephants, cetaceans; what's amazing to me is that what's known about the network mapping of these individuals, of these species, is that those networks look incredibly similar. Elephant networks and primate networks and dolphin networks look very much like ours.

To me this begs what is a really interesting question, which is, maybe there's only one way to be social. I mean, why would it be the case in the natural world that whenever we go looking at social species, leaving aside the eusocial insects, would they evince these network properties? Because the last common ancestor between us and whales was 60 million years ago. So whales clearly have evolved independently, and with elephants, it's about the same.

So, they're converging, by convergent evolution on a similar solution, not on a bodily phenotype, but what James and I are calling an exophenotype. So think about this (to borrow an example from Richard Dawkins): Why is it that if a spider evolves bigger mouth parts to capture more prey, we think of that as a kind of evolutionary adaptation; but if a spider evolves the construction of a more elaborate web that basically achieves the same thing, we don't necessarily think of that as a phenotype? Well, actually we should. Let's start thinking of it as a phenotype—spiderweb morphology is a phenotype. If that's true, by a few short leaps I could get you believing that social network construction is a phenotype. My manipulation of the social world to construct the network around me, I would argue, is no different than the spider's manipulation of the physical world to construct a spider web around it.

Second, picking up on your point, what's amazing to me is that dragging with it, not necessarily dragging with it, but walking along with the network structure are all these other things. For example, mirror self-recognition. Dolphins have mirror self-recognition, primates clearly do, elephants have mirror self-recognition, cooperation, self-identity, and other-identity. So if you're going to cooperate and form networks with non-kin, you have to be able to know "Oh, this is June, and that's Sendhil, and that's Danny..." you have to know who they are from moment to moment. And these other animals also do that. So there's this suite of features that seems to be necessary and go together for the construction of social worlds.

SANTOS: Just to follow up, do you have to be the kind of cognitive creature who could do X, Y and Z and then you're like, "Oh, I'll talk to June and then, ooh, the network forms?" Or does the network form and that creates this crazy selection pressure to have these mechanisms ...

CHRISTAKIS: Yes, it's both, I think. That our social life and our biological heritage are in a conversation across eons.

Think about this: Imagine a beaver, for whatever reason, has a chance mutation that makes its behavior different so it constructs a bigger dam. And now when the beaver constructs a bigger dam you get a bigger flood behind the dam. Now across time those beavers, ideally to exploit the greater linear perimeter of the pond that they've created, which gives them more forging opportunities, need bigger lungs. So the beaver now, because of the behavioral change, has to start evolving bigger lungs to be able to be underwater more to explore this perimeter, or bigger flippers or whatever beavers need to be effective. Okay?

Well, I think humans are like that, actually. We have little things where we begin reworking the social world around us. That creates selection pressures on our brains and our cognition; it makes us social. The more social, cooperative, mirror self-recognition, all that other stuff we do, the more able we are to create these webs around us, and it feeds back on itself. But what's so interesting to me and James about the social world is that, unlike the physical or the biological world, which is 'God-given' [or exogenous] and all around us, we create the social world. We create the selection pressure that then feeds back and contorts our minds and contorts our bodies. That's what we think is happening.

GRUBER: You're talking about social contagion and I know earlier we talked a little bit about emotional contagion. I'm just wondering to what extent do you think the spreading of this phenomena is going to vary depending on the type of network we're talking about, whether it's specifically, I think of offline or in vivo interaction and now with the social media networks growing and ever-increasing, and the degree to which you're expressing emotions in these two domains is radically different.

CHRISTAKIS: One of our arguments has been that, with respect to emotional contagion (in which we're very interested), there has to be some relationship at stake. My emotional response to my child in pain or my colleague in pain even, depending on the colleague, is very different than my emotional response to a stranger in pain. I still have empathy, I'd like to believe, and sympathy for the stranger in pain, but there's clearly something different about it. Plus, it's also different to see the person in pain than to read about the person in pain. So, (a) the nature of the social tie, and (b) the visibility, are crucially important. However, I think that you can transmit emotional states to a lesser extent, but still, through online interactions. Like if you get a sad letter from your sister, you're going to feel sad about it even though it's a printed word and not quite as powerful as seeing your sister.

We have an unpublished paper, which I can talk about very briefly, in which we exploited weather variation as an instrument. We looked at all the residents of New York City, and if it rains in New York City, with Facebook mapping of the whole country, their Facebook friends in cities outside of New York are affected by the weather in New York City, to two degrees removed. I won't go into all the details, but we did this in the econometrically way, and so we can discern, in a kind of quasi-natural experiment, to the extent that you believe the literature that weather affects people's moods (which there is a nice cottage literature on this), you can use that as a kind of, what's known as an instrument, to identify these effects between online friendships.

KURZBAN: Like all models, there's a certain degree to which you're abstracting, and that's a necessary feature of modeling, right? You've got to take some stuff out. And you highlighted something that's right about psychology, which is that we don't spend that much time thinking about what friendship is for. There's been some assumptions about it being for exchange and so on, and you have a different proposal.

Just to connect you back to some stuff that we've been thinking about, one thing that seems to be important in our data about friendship is that the nodes aren't equally weighted. The amount of time and the degree to which I'm close to my best friend is really different from my fourth and fifth friend. What I'm really curious about is, first of all, as a technical matter, how easy it is to build things like that into the model. Just for the record, my suspicion is it's going to be really important. It might even change your packaging data.


KURZBAN: Because if I spend a lot of time on doing that … Right? So, as a psychological matter, that seems like a reality, which would be very cool to build into these sorts of things. And, again, as an empirical matter, we're finding that there's a relatively nice function that one can use to map these things. So, is the future of this kind of weighted edges ...


KURZBAN: And what's going to happen?

CHRISTAKIS: Yes, there's a big move to weighted graphs exactly for the reason you're describing. Every tie can get a weight now, so you can describe ties and not just the nodes. In fact, ties can become just as complicated as people. How long has the tie been lasting? How intimate is the tie? How frequently do you see the person? What's the vector? Do I say, "You're my friend" or you say "I'm your friend?" And so you can begin to have all kinds of details, which are highly relevant, and you can weigh the ties and use a variety of methods which allow you to take advantage… And it falls mostly as you would predict, right? So, just as you suggested, people with whom I spend a lot of time are more important paths through the network when it comes to germs, for example.

BROCKMAN: Unfortunately, Wallace Stevens couldn't be here today, but he asked me to read the following excerpt from his poem, "United Dames of America":

"The mass is nothing. The number of men in a mass
 Of men is nothing. The mass is no greater than

 The singular man in a mass. Masses produce
 Each one its paradigm."



nicholas_a_christakis's picture
Sterling Professor of Social and Natural Science, Yale...


sendhil_mullainathan's picture
Professor of Economics, Harvard; Assistant Director for...
june_gruber's picture
Assistant Professor of Psychology, University of Colorado,...
fiery_cushman's picture
Assistant Professor, Department of Psychology, Harvard...
robert_kurzban's picture
Psychologist, UPenn; Director, Penn Laboratory for...
joshua_d_greene's picture
Cognitive Neuroscientist and Philosopher, Harvard University
laurie_r_santos's picture
Professor of Psychology, Director, Comparative Cognition...
joshua_knobe's picture
Experimental Philosopher, Yale
david_pizarro's picture
Associate Professor of Psychology, Cornell University
daniel_c_dennett's picture
Philosopher; Austin B. Fletcher Professor of Philosophy, Co...
daniel_kahneman's picture
Recipient, Nobel Prize in Economics, 2002; Eugene Higgins...
Paperback [ 2011 ]
James Fowler
Hardcover [ 2013 ]
Daniel C. Dennett
Hardcover [ 2013 ]
Joshua D. Greene
Hardcover [ 2011 ]
Daniel Kahneman
Hardcover [ 2011 ]
Robert Kurzban
Hardcover [ 2013 ]
Sendhil Mullainathan, Eldar Shafir

Conversations at Edge