When we look at your career trajectory, we see that you came from Transylvania and ended up in the United States. You move constantly between Boston, Budapest, and Transylvania. Can we argue that you live your life in a peculiar network of places and cultures?
The way I often think about it is: my home is in Transylvania, I live in Budapest, and I work in Boston. These three places have different meanings and define different patterns for me. I'm mostly in Boston for work, where my social interactions are limited to colleagues and professionals, and even though I do have friendships and stronger ties, they all come from people I have worked or am currently working with. In Budapest I have a much more heterogeneous social network. That's where I go to have fun — fun beyond work. That's where I do much of my art. The people I work with in Budapest are mostly involved in the art world — young (and not so young) artists who help me develop the ideas and the artworks that we make. And of course, home is Csíkszereda, Transylvania, where I'm from and where my mother still lives. My wife occasionally says that I totally change my behavioral patterns in Transylvania. There's no place where I can relax and turn off as I do there. I behave and relate to people very differently in the three places. I do like heterogeneity.
As an editor and literary critic reading your work, what struck me first is how you use stories and anecdotes to explain complex phenomena and abstract scientific ideas. Do you have an anecdote about how you realized the importance or relevance of networks?
Let's talk about storytelling first and then about networks. I grew up in a family that was not related to any technical field at all. My mother was a theater director, my father a historian and writer. When I started university in Bucharest, I worked at the magazine A Hét, a weekly national magazine in Transylvania in the 1980s. I was a writer before I was a scientist. And an artist, because before I discovered physics, I was preparing to be a sculptor. This love for writing and clarity has stayed with me. When network science came around and I ended up writing my book Linked, I was not starting from scratch. I had a solid background in storytelling across different magazines.
How did network science start for you?
For me, the story started in 1994, when with a fresh Ph.D. I became a postdoctoral researcher at IBM T.J. Watson Research Center, just north of New York City. It was a prestigious ivory tower where the best graduates around the world were hired and given the freedom to work on a wide range of subjects. I didn't really have a subject prescribed to me. Working at IBM and living in New York City at the time, just before Christmas, I decided borrow a book from the library to learn: What exactly does IBM do? What is computer science? I graduated with a Ph.D. in theoretical physics, so I had no idea what computer science dealt with. It was in that book that I read over the Christmas break of 1994 where I first encountered the concept of the graph, and the concept of the network, mainly in the context of electronic circuits. But living in the Bronx and in New York City, I realized there were so many other networks around us that kept the city working, from the pipe that brought you water into the apartment, to the electric wiring, to the phones, and even to the emerging Internet. And I thought it was so interesting that no one seemed to be studying these phenomena, how these networks tied the whole city together and made it functional, and how they worked. During that break I decided to start studying networks.
Did you know at the time about the Hungarian roots of network theory?
At the time I had no idea what anybody had done on networks. It was more of a curiosity on my end. Within days I discovered that some of the most consequential advances in graph theory were done by the Hungarian Mathematical School, by Paul Erdős and Alfréd Rényi, to be precise, and more recently by László Lovász, Béla Bollobás and many other mathematicians. It turned out that Hungary had a very strong graph theoretical school, focusing on random graphs. And that’s crucial because the networks around us are not regular objects. Think about social networks. The question of who our friends and acquaintances are is driven by lots of apparently random processes. You meet somebody at a party, you get assigned to the same room in the dorm, etc. Lots of different events that define whom you know and whom you're close to appear to be random. Yet society is not random. If it were random, we would not be able to function. There is an underlying order in society as a system that allows us to help each other, to communicate ideas, to pass on information, and so on. And they more or less work. The challenge was: How do we go from these microscopic, random processes to what we call macroscopic order? That thinking was, in the quantitative way, started by Paul Erdős and Alfréd Rényi, who imagined networks to be purely random, that is, the dice decides who's going to be our friends. It was through reading them and understanding the mathematics that reinforced, for me, that 1.) networks are an interesting problem and 2.) we know very little about them. Because the very strong mathematical results remained a hypothesis that was never matched with reality. I'm a physicist and physicists are good at math, but equally good at thinking about whether something connects to the real world. Physics, fundamentally, is an empirical science. For a mathematician, an interesting mathematical problem is worth pursuing on its own. A physicist finds a mathematical problem interesting only if we can connect it to the real thing. And that has been my journey in networks: understand what real networks look like, what laws govern their emergence and evolution, and developing the mathematical concepts to describe the processes that we see in the real world.
Could you briefly explain what the underlying principle was that you encountered in different networks?
To understand that, we must go back in time to see what mathematicians thought these networks looked like. Paul Erdős and Alfred Rényi, in 1959 and 1960, ended up writing a series of eight fundamental papers motivated by real networks. And they said we do not know how these networks were wired, but they surely seem to be random. So let's assume that they're truly random. The dice decides who's going to be your friend. They built a theory of what we call random networks. People think random is not predictable. In a scientist's mind random is predictable because there are certain aspects of randomness that we can test. One of the predictions they made was that if you looked at how many links each node has in a network, then these random networks would be very democratic. That is, most nodes would have a comparable number of links, and it would be rare to find nodes that have ten times more or ten times fewer links than the average node. And the bigger the network, the more similar the nodes would be. For example, if society truly were a random network, given that sociologists have estimated that the typical person in society knows about 1,000 people on a first-name basis —the most popular person would know only 1,150 individuals and the least popular about 850. In other words, in a random world we are very much alike.
But we are not.
Not at all. We discovered in 1999 that this was not the case on the World Wide Web. There were a few major web pages (like Google and Facebook today) that had hundreds of millions of links that coexisted with most web pages that had only four or five. The same pattern was repeated in cells: most proteins only interact with one or two other proteins, but a few interact with hundreds. And the same was true in social networks, like with Twitter followers. Most people have only a thousand or a few hundred followers while a few have tens of millions. The existence of these highly connected nodes could not be explained by existing theories. And what we showed in ’99 was that these highly connected nodes and the underlying architecture of these networks, which we named their “scale-free” architecture, could be explained by two very simple principles. The first was that networks don't emerge fully formed, but rather they grow one node at a time. Every network out there is the result of a growth process: sometimes it took twenty years to arrive at the current network, sometimes, like in the cell, four billion years. The other principle was that when new nodes arrive at the network, they don't choose where they're going to connect randomly but have a bias towards the more connected nodes. That’s what we called “preferential attachment”. And we showed mathematically that growth and preferential attachment can explain the scale of the architecture. They can explain the emergence of the hubs, predicting the observed number of the hubs, and how they acquire links in time etc. And that was a major transition in network science because suddenly we started to have models that not only were able to predict the behavior of a single system, but could be, and have been, applied to many systems.
This is why network science is universal.
Exactly. Universality means that we don't have separate theories for social networks, biological networks, the Internet, and economic networks, even though these are very different systems. In network science we have one framework that can describe all of these. And of course, there are particularities to each of these systems that are interesting and worthwhile studying on their own. But it’s a bit like building rockets or airplanes or slingshots, which all follow the same fundamental, universal law of gravitation. Even though they look very different and the process of using them is very different, you first must know gravitation to build them. Only after that can you start talking about the details. The way we think in network science is that, yes, the social network is not the same as the world wide web or the cell, but they are governed by the same universal laws.
In 2020, COVID-19 made the global population more aware of the existence and importance of networks. Did COVID teach you anything new about networks or was it just a new example of a structure or algorithm that you were already familiar with?
About ten years before COVID, I did a documentary about viruses and their spread in networks. I specifically claimed that it was not a question of whether we were going to have the next major pandemic, but rather when and how deadly will be. I said it because there is a twenty-year history in network science of studying potential epidemics. Even my lab published a paper ten years before COVID where we used mobile phone data to ask how a potential virus spread by physical contact would spread around the country. And of course, that potential virus happened a decade later. There is a discipline that we call network epidemiology initiated by my colleague Alessandro Vespignani that focuses entirely on this. What’s more: already before COVID, they had very accurate tools to predict, given the characteristics of the virus, how many people would be infected, how fast, how many in Budapest, how many in Boston… And because of that, already in December 2019 we knew in the lab that there would be a massive breakout. I remember talking repeatedly to Alessandro, who was running these models, and he was concerned about who we needed to tell, particularly in America because at the time no one wanted to listen, particularly the leadership. This was the Trump administration. When they were finally willing to listen, Alessandro became the White House modeler on COVID, and he continues to do that job even today under the Biden administration. The network science community was prepared for COVID. We knew it was going to come in some way or other, because transportation and globalization have reached the point that there's no way a virus can emerge in Southeast Asia or Mexico (like H1N1) and not spread if it has the right characteristics. The question is how fast and what we can do to stop it. COVID was transformative for network science because terms that we only heard at our scientific conferences, in our scientific papers, became dinner table discussions. People started to talk about social networks, connectivity, mobility, reproductive numbers, and so on. So let's put it this way: COVID has really helped the world to understand the important achievements of network science over the last twenty years.
So the phenomenon of COVID taught the world much more about networks than it taught network science.
That's right. Network science became mainstream. Even though people didn't know they were doing network science, that line of thinking has entered everybody's mind and is still there. The discipline has gotten a degree of recognition and visibility that under normal circumstances would have required another 50 years to reach.
How do you see the future of pandemics?
Oh, that is easy. During COVID, every government swore they were going to commit resources so something similar could not happen again. Give it another ten, fifteen years and they will forget about it. And then, say, fifty or a hundred years from now, when the next virus will come along, we'll start over. But on the other hand, I should also say that this has been the first pandemic where absolutely everything has been recorded. We know who got sick, when they were infected, who they were in contact with, etc. It's a pandemic where we have an extensive amount of data. I think we're going to be much more prepared as a scientific community in how to approach the next pandemic. That knowledge base is not going to go away. But I'm pretty sure the governments’ willingness to invest in epidemics will go away with time. I knew of a company who had used Alessandro’s tool to build prediction methods for a potential pandemic, and the company folded six months before COVID because they could not get any investment since no one thought it would be a problem worthy of study.
Murphy’s Law.
Exactly.
If I understood correctly, the main goal of network science is to visualize or chart the unknown and the highly complex. There is probably no phenomenon so unknown, chaotic, and complex as the future. Could you give us an example of how network theory can be used to imagine a possible future?
First, the future is easy because it will arrive predictably. The problem is the idea of “predicting the future”. People think about it in a sense of what the course of history will be, who the next Nobel Prize winner will be, or who will make a big discovery. We scientists think about future prediction differently. Future prediction is about trying to tell how many people will get sick given the properties of a virus. Trying to tell how long the epidemic will last, what age group will be affected and how we can intervene. Those are all predictions about the future, about sorting alternatives and providing tools to intervene and change that future. Very few scientists are engaged in saying what's going to happen 200 years from now, but many scientists are engaged with what the temperature of the Earth will be if we don't do anything about emissions, and what technical capabilities we have to change it. Scientists are also involved in trying to tell us if we need to put masks back on again. Now COVID seems to be over, but quite a number of viruses are circulating, and what would be the impact of, say, 30% of the population wearing masks? How many lives would be saved? So, on the one hand, there is the pragmatic aspect, where science and engineering are involved predicting the future. Even when you design a bridge, you're trying to predict the future, as you're imagining how it will be used and how we should design it so that it will not collapse. On the other hand, we are asking whether we really can predict complex phenomena that have not happened. I just have to go back to my childhood to the science fiction books and their predictions about what would happen by the year 2000, that we would all have individual flying bots, which we certainly imagined in the 1970s and early ’80s. Instead, what we have is an iPhone and access to any piece of information, anywhere, anytime. And we have ChatGPT and AI that can respond and have a conversation in a way that is indistinguishable from a conversation with a human. It can correct my text in a way that looks like it's written by a professional writer. None of that was on the horizon. We therefore have to define what we mean by “predicting the future”. It's about predicting certain phenomena that are essential to humanity, like global warming, environmental effects… or imagining new things that could come along. The second part — no one has a clue how to do that. But for the first part — science is doing an increasingly good job.
One could imagine network science as something dealing with dots and lines — you just need to connect the dots. But if we say that the dots and the lines are already given elements, and a certain network might only work if the particles or elements are already established, then what if we introduce a new element, which could potentially rearrange the whole system, its order, and essentially how we classify its elements? Could we argue that this newly introduced, unknown element is what represents the “future” in the equation?
You are right. One of the inspirations for me as I was developing network science was Isaac Asimov's Foundation Trilogy. The underlying story of Asimov's book series is that humanity reaches the point where there's a group of people who develop social mathematics that can foresee the future hundreds and thousands of years in advance. But there's a price for that. They can only do so if there's no innovation. To maintain the future they predicted, they have to suppress all innovation. The crucial narrative twist is the emergence of The Mule, who changes the future by altering the patterns. Asimov — who himself was a biochemist at Boston University— saw that if you want to think about the future, you cannot allow new elements to enter, because science is only good at predicting the behavior of a system, simple or complex, if all its components are known. Once you throw a new component in, the system becomes unpredictable.
Some say that one of the most difficult parts of network science is data visualization or the modeling of complex systems. You have recently shown a new aspect of your creative personality by exhibiting visualized data in art galleries. While I would leave art critics to decide if these works are more related to art or science, I am more curious about how you see the relationship between these two spheres. There is a long tradition that says that art ventures into the unknown and might irrationally create images or patterns that can be used later for analysis.
I always think that fundamentally the way an artist and scientist think is very similar. The only difference is that scientists must match their ideas with reality, and artists don't have that constraint. That freedom is exciting as well but the main reason I'm in the arts is that I'm a visual thinker. I finished both my undergraduate studies in Bucharest and Budapest, and my graduate studies in Boston, without ever going to class. And the reason I didn't go to class is that I could not retain the knowledge given to me. I can retain knowledge that I read, draw, write down myself. I developed a visual way of doing science— I had to draw it or show it somehow. When I became a professor and had a lab, I started to use this visual gift not only to develop the mathematics of networks but to visualize them too. For me, science and visualization were part of the same intellectual journey. The very first product of my journey into network science was not a research paper but rather what you would call a visualization, where I used the computer code to build the network. Today, 30 years later, it is called generative art. I used the computer to generate that image over and over. Right now it's being exhibited in Budapest in a gallery where our computer generates another image every second using the code I built in 1994. Over the last 25 years in my lab, we had three parallel ways of sharing ideas. The first was the research papers that were hidden in scientific journals and relied on a scientific jargon that directed its context to the scientific community. The second that was much better known to the larger public were books I wrote addressed to a general audience, in which I tried to summarize our findings and put them in context. But there was a third journey that was seen only occasionally, which is the visual story. Seeing what I was researching helped me to write about many things in a much more credible way. But there were no venues to show these visualizations. This all changed five years ago, when museums and galleries started to discover those and to exhibit them. You may call them visualizations, I call them network art or data art, to be more precise. They are part of an emerging movement in art that I generally refer to as dataism because there are many other artists who rely on data. And most importantly, I call our process inverse appropriation. What do I mean by that? As you hinted, many artists have a common practice of appropriating forms and ideas from other disciplines. There's a long history of artists taking objects and images from science and putting them into an artistic practice. Ours was the opposite journey. I was first studying to be an artist before turning to science and I even took art classes parallel to my teaching at the university. I started to bring the modes of thinking about art to our scientific practice, to represent the work that we were doing. So, thanks to inverse appropriation our work has now reentered the space it came from: the museums and galleries. In the last few years, we have had a very busy exhibit schedule, showing our work from New York City all the way to Germany and Milan, for example.
Is it somehow related to what's happening with artificial intelligence? Because right now we are just bombarded with artistic images generated by AI But there are a lot of concerns related to that. What's your stance on that?
Well, AI is transformative because it has demonstrated that it can make very sophisticated images very fast from just textual input. Yet, the speed and volume cheapens art. We are deluged by these images and they're so recognizable that on their own they have ceased to exist as an image of value and more as proof of what AI can achieve. This is a moment like when photography came along. Before photography, the pursuit of art was perfection in representation. Art was slowly edging towards photorealism. Once photography arrived, the Paris Salon pretty much collapsed. The classical way of doing paintings was over. Impressionism emerged because there was no need to do realistic painting anymore — photographs could do a much better job. We have a similar moment right now with the emergence of AI as many of the things that artists could do with great effort in the past are done in a fraction of a second by the AI. The question is, what is the role of the artist as we go forward? It's clear that AI will become a tool in the artist’s journey, just as photographs and photoshop did. We will find a way to go forward and incorporate this ability into artistic practices. And the challenge will be to use this new tool in a meaningful manner and show a path forward for the next generation of artists.
Coming back to the topic of networks and the future, what struck me while reading your work are the principles of distribution within networks. As you hinted earlier, specific websites get all the visitors while other websites are basically invisible. Specific people have vastly more followers than the rest. If we argue that this principle of disproportionate distribution is only enhanced by technological development — the way giant companies like Amazon, Googleand Facebook are in a hegemonic position — the future doesn't look so democratic.
I certainly agree with that assessment. Inequality has been growing since the Second World War and doesn't seem to be stopping. That aspect certainly needs to be fixed. We need a much fairer tax policy and distribution of wealth. There's no question about that. Sadly, it is wars that redistribute the wealth. In times of peace, there has been a constant sliding towards inequality. If that is indeed true, then this is a consequence of a prolonged period of peace. But we should wake up to that. And there are tools out there, from tax policy to redistribution, that would allow for interventions to achieve a more fair world. This requires societal will. People need to show their voice, to say: the way this is happening right now is not good for us. Change doesn't necessarily require a revolution. It could be done with relatively small changes to the tax code for example.
One topic impossible to ignore when talking about the future is its ecological aspect. Due to the ecological crisis a large number of people don't see the future as a horizon of open possibilities, but more as a looming threat. How do you relate to that and how do you see the role of the sciences in this situation?
I've never been a Luddite. I've always believed that humanity can avoid catastrophe. We're lazy, so we do so at the last minute. But we've been good at that. When nuclear power came along and it was first manifested in the form of the bomb, everybody was convinced that the earth will end in a nuclear war. Luckily, we avoided that. And I strongly believe that we will continue to do so. The same is true, for example, for biology. We have created tools that, if unleashed, could easily wipe out humanity or most forms of life on earth. We have not unleashed them, either accidentally or purposefully, because we are aware that these are dangerous things, so society regulates the forces that could unleash them. We tend to be careful in that sense. Global warming and the ecological disaster we are facing are of a different nature because we are causing it by many, many small actions. And ending it requires global action and there's not a single authority who can achieve that. It would require a massive concentration of will to induce change. But we see the signs of some agreement. I'm not saying they are 100% pursuing it but they're edging towards it. I personally think that the answer to global warming will be technological and scientific, rather than political or social. I even see discoveries on the horizon that may turn us around. We have seen in the news in the last two weeks the advances in fusion. What the news has failed to talk about is that there is a startup company here in Boston with more than $2 billion in the bank, promising that by the end of next year they will have a working commercial fusion reactor. It's unbelievable. Let's say that this will not happen next year, but five years from now — once that happens, it will be the end of fossil fuels because we could generate a pretty much unlimited amount of power from water. Yes, it is painful to watch the unwillingness of our leaders to come together to decrease the use of fossil fuels, invest more in renewable technologies, and so on, but I'm hoping that once again science will save us from the incompetence of our leaders.
Are you more optimistic or pessimistic about the future?
As a person, I have always been optimistic that we will find solutions. Pessimism doesn’t help us go forward. It may be a valid state of mind, particularly with my Hungarian background, where pessimism is a national sport, but I always try to be optimistic.
It was originally published in The Continental Literary Magazine.
