Focus 580; Linked: The New Science of Networks

Good morning and welcome back to focus 580. This is our number two of the program it's our morning talk show My name is David Inge. Glad to have you with us Jack Brighton is the producer Harriet Williamson helping out this morning with production Henry Frayne at the controls. A quick mention that tomorrow morning on the show in the 10:00 o'clock hour we'll be talking about how computers are increasingly now being used in the social sciences and the humanities at our guest for the program will be Orval Burton. He's professor of history and sociology here at University of Illinois in the second hour of the issue is gun regulation and public health and look at the issue of guns as a public health matter. We'll have two guests Sue Paschen who is a firearms project manager at the Consumer Federation of America. And also Brooke Anderson a gun regulation organizer with a local organization here the Champaign County health care consumers. We're here weekday mornings from 10:00 until noon and we have 10 different topics every week. We always welcome your questions and comments. In this part of focus 580 we'll be talking about the new science of networks.

And our guest is Albert Laszlo Bara basi. He's professor of physics at University of Notre Dame and for the past couple of years he and his colleagues there have been doing research on complex networks. He has written about some of what they have found in a new book that's touching off a lot of interest. The title of the book is linked the new science of networks and I think the subtitle really says it all. That is how everything is connected to everything else and what it means for science business and everyday life. Perseus is the publisher of the book and in the introduction to the book he explains that the book has a simple game to get you to think networks. It's about how networks emerge what they look like and how they evolve. Networks are present everywhere he writes. All we need is an eye for them. And what he and his colleagues have found is that. Networks in the real world function rather differently than scientists had previously thought as they constructed

models thought about connections between the various elements of networks. They thought of them as being random and in fact they're not. He as colleagues have discovered that they definitely follow rules and that you could see them operating in all kinds of situations from everything from the global economy to the Internet to the way diseases operate. So as we talked this morning with our guest. Questions are certainly welcome. The number 3 3 3 9 4 5 5. We also have a toll free line good anywhere you can hear us 800 to 2 2 9 4 5 5 questions are welcome. The only thing we ask callers is people just try to be brief. Just so we can keep things moving along. But anyone who is interested in calling is welcome to join us. Hello. Hello. Thanks for talking with us that was this year to begin I think to make certain that people know what it is we're talking about. We should just talk in very basic terms about the concept of a

network. And they have a very basic definition. What is it that we're talking about what is a network. Sure. The basic definition is anything is a network where you have a bunch of notes connected by something which we typically called links but the best is to restock with examples you know if you think about the society the society is nothing but the network of people connected by friendship acquaintance links or professional links you know so if I know my friend there isn't in the building between us essentially meaning I can always pick up the phone and call him and you know ask for a favor. But you know if you think of the war society we really have a very very large net for it because every person has about like thousands of acquaintances that therefore is connected to a thousand people. But there are many other examples of networks I mean if I think my body is you know we always think of being nothing more than a bag of chemicals but we wouldn't. But those chemicals would not be really able to create life is debarred and would not interact with each other. So I can go ahead and throw a map off the cell for example for every node it's a

chemical and every reaction that connects to chemical would be a link. The board why do we have you know essentially the basic big online library is another network where the notes are the best pages and the links are how you know until you are ads that you can click and go from on that page in another one and really the reason why it's so wonderful these days of war that is because of Billings because the information is connected to each other and not just stand alone information and finally to get a final example would be the reason you can understand what I'm saying is because our language is a network of warts the warts are connected to each other by meaning by various syntactical relationships and due to these links are the ones which make you know from just a dictionary before it's something that you know at least communicate make make communication possible make language. So before the the work that you and your colleagues did as I understand it as people thought about networks that is these collection of points or nodes and the

links between them they think as they made up models and they thought OK well it will start out we'll construct a model network will think and then we can think about how that works. They made the assumption that the links between the nodes were random and they as they built their models they would they would put on put nodes on a piece of paper and they would make links between the nodes and then they would look at how many links each node had and some would have a few and some would have more. And then if you plotted them out on a graph the number of links particular nodes had assuming that that's the way things worked that you would end up with the classic bell curve distribution and that the big sort of the big breakthrough that that you and the people who work with you discovered was that in the real world it's not like that at all. That's correct. And just to step back. The beauty in his story that random that broke more

was due to doing the work of two hungry a mathematician. Poor speech were for many many of your audience probably heard about him he will be very eccentric mathematician who would leave out of the suitcase and travel from one the university did not there was stay a few days do some PR and then move on. And he's a colleague and friend of ours for the rainiest So they were the ones who really proposed and the 1050 is deceased. That's work are so complex and you know the extra day use the example of communications and and living systems and the networks appear to be so different from each other. The simplest way is to describe them used assume that they are random and as you said one of the consequences if I randomly put the links is that I will have about curves. Now why why is this about could be interesting there is one point that I would like to emphasize here. A backer is really telling us that in a random network most nodes have approximately the same number of links. It's really a mathematical formulation of the fact that it's

very rare to find a node that would be very very dirty compared to the average node so we do group a good approximation you can assume that just about all of those have roughly the same number of links. When we started looking into networks at the end of nine thousand eight ninety eight beginning of 99 where we did it we wanted to find an example for a large network that we could study and we started with the war divided by AB and essentially. What we did is that we wrote the bits of software that went out and created for us of that build toward that by going from page to page and return just maps to us and indeed as you as you mentioned that there was a big surprise in that result that this Democratic Network was not present at Tor you know that we expected that most North would have roughly the same number of links on the ward live but in contrast to that what we found is that the majority of the pages on the board Live had no one or maybe two links to them. But there are a few pages that have thousands and thousands of them. Now D.C. extreme distribution in

particular its existence which is very very highly connected No it's simply could not be consigned Mitterrand on that record to view but we really saw that on the war divided bad data a few hops hops being is very highly connected nodes that keep the one that work together and some of the top stock I mean most of these hops are actually familiar to us you know they are Yahoo. Some of the big news organizations that have so many links to other sites and so many people are linked to them that day they appear at major hubs in the ward live but there was only one part of the story that we found out in the void by the cops. The real surprise came. Then people started to look. We're not there for to look at other networks. So for example you can create a network of Holywood actors by collecting connecting any two actors together they play together in a movie and somewhat Your audience may be familiar with the Kevin Bacon game or so degrees of Kevin Bacon. But again I asked a participant to connect any actor to Kevin Bacon. Wired to movies in which they played together. So behind the game

is really a very large very very densely interconnected network what to Holywood dearly. But when we looked at this network and we can start doing that for instance 1892 today we had all the actors and all of the movies in there. What we found is that most actors had links to maybe five ten other actors so they played in one movie to fight other actors together. But there were a few actors that had talents and thousands of links. So again the same pattern emerged that there were a few very very highly connected hubs that kept the network together kept Hollywood together and these examples have beauty of curing in the last few days in many other networks. You know we looked at the cell how chemicals are connected to each other by reactions and again we found a few chemicals that have a very very large number of links that appear to be the hops of the cell a group a group of scientists from Boston USA tennis talk home has analyzed the data how people are collected by sexual relationship to each other so always putting a link between two people if they had sex to get there and what they found that in this network

most know what studies. People had you know five to 10 buildings there and lifetime meaning that they had five to 10 sexual partners during their lifetime. But there were a few people who had a house and some of them. So no surprise but that in the last few years we understood that this very homogeneous structure which we kind of chord scale free networks gift network meaning that there's this great distance of just very very large and it's many many more want the pure in just about any large network that is out there so it seems to be a common architectural complex systems. Our guests let me introduce again just very quickly is Albert Laszlo Barabbas. He's professor of physics at University of Notre Dame and directs research on complex networks there for several years now the past few years. He and colleagues have been doing work on networks. And if you're interested in reading about it you could take a look at his book which is titled linked the new science of networks. Also he has a web page that I found just by putting his name into the

web browser. But you can also go to w w w dot and D for Notre Dame and the dot edu slash Tilda networks slash and you will find it and there there are things like articles that he and his colleagues have written. Talks they have given news articles about it. So if you're interested in reading on this subject you can go there to that page and look at various kinds of material and of course you can also head out to the bookstore and buy linked if you would like to read about it. And if you'd like to ask a question of course you can call them 3 3 3 9 4 5 5 toll free 800 2 2 2 9 4 5 5. So you discovered that this basic idea that previously the assumption had been that that networks were the links in networks were random and that what you found out first by studying the web and then looking at other kinds of networks was in fact that they weren't that way at all that most nodes most of the

points in the network have a small relatively small number of links but a few of them have a whole lot of links and that they. That that's true of the internet but it's also true of awful lot of other kinds of networks can be. And we conduct by communication networks and economic networks and even biological networks. So then I suppose another question is well it seems that networks organize themselves that way. We see that that pattern in all kinds of networks. So a natural question would be why. Why are networks like that and why are they not random like we thought perhaps they were. That's right that was obviously the next logical question and took us a few months actually and we kept thinking about it what why could that happen. And then I know you would realize you know that we have to start from this very simple assumption of that networks are on them and try to think what these Vong would def assumption I mean each part of the solution may be spilling into the system

and one of the houses that goes into just random that record of you is that if I want to more Dylan at work let's say of the ward allied lab and that has but two billion notes which approximate the number right now. Then what I should do is to go ahead and write to my that's a wide 43 billion two billion notes and only connect them. This may not seem to be a particularly big assumption but very very thin assumption hidden in it that is that I have available for me all the notes in the more that I want to mark the moment that I want to look at it to proof is that the networks never popped up with three billion notes if you think about the war divided in 1991 it had only run about page 18 burn us leave or die to create terrorists. First by page and then people were discovering the media and they were adding new pages one by one and essentially the network grew out from Rampage to have over two billion pages today. And so essentially what we look at look at the world what we find is that the network emerged by growth but that pattern is present in just about

all networks if you think about it in Hollywood you know in 1890 there were only a handful actress playing in silent movies and a new actress joined a trade as Hollywood became bigger and bigger. Network slowly increase from a from a dozen actors to today over half a million actors and if you think about our But ourselves you know in the region only about 3 billion years ago we just had a few chemicals going on and only on Earth and the prevailing view is that the sense within you a new comic opera sympathised and the chemical network has grown to become a cell and to become a larger and larger and more and more complicated. So if you think about evolution that's a growth process from a few chemicals to large a larger and more complex molecule and just about any network out there if you look at them you will find a spectrum that they started out with a few notes and they grew out of them and became what we see today. The growth is a very essential property of networks and we should not more than that for such that we assume that we have all the notes available. That was

one thing which was missing from the random network model but there was another a key element which was necessary to explain the hops that is when you know a constant the network the random wars you assume that it connects and only to any of the nodes to power out there if you think about your network that is not the case. When you create that you've got pigeon you would like to connect some very you know let's say to some news outlet you would not pick any pitch out there that would have been used in it you know you would go to the pages that you are familiar feats and you know you go to New York Times CNN or salon or NPR and what you would what you would really connecting to are the pages that have a large number of links in the first place because you do without the pages that you are familiar with it you have to realize there are two billion pages out there on the ward 5 and they think a person is familiar with that tiny fraction of them but most of us are familiar hops with the highly connected nodes. So whenever we decide to connect summer we connect to decide that wow familiar of it and that includes an immediate bias towards the more

connected nodes because those are more visible we are more acquainted with them. So you know it's on the war Dr. Bethany creates a new link it does not create deafening randomly but it's peace first to connect to the more connected nodes doesn't always get sometimes to connect to our friends but which may not be that connected but. In general there is a preference this is what we call preferential attachment towards the more connected nodes and the same pattern is present again under not to excess Well you know in Hollywood if you are if you are an actor that has a thousand flings more likely and a couple other actors who would be chosen for another movie because simply you are visible you are not to do with DSN directors want to employ us actually to bring in the audience. So if you look around in just about any network you will find that there is always a bias to connecting to the more connected nodes. The bias in from one it took another one may have a different origin but the bias is present. So now you have a network which is growing every day always and you know it is coming again and the new notes tend to connect prefer to connect to more connected notes

and what you find if there isnt more highly connected node out there that they connect to that we grow faster than the less connected nodes. So there is what we call the rich get richer phenomena don't reach notes acquire more lings during time because they are more of a support and will leave very much behind the less connected less rich nodes and this mechanism is the one which seems to be responsible in most of the networks for the emergence of the hops or the very very highly connected nodes. What is interesting though is that the word kill for not for guitar used to describe these demented networth also implies that there is not a single hub in his networks. There's always a hierarchy of Hobbes. There is a most connected know it that say that would be Yahoo right now and develop and there are a few that are very close to it you know Google and some other search engines that are almost as connected but not quite. And they are even more that are a bit less connected and if you go through the rest of us connected notes you find more and more of them. So when I think that there are

hops in these networks and they are merciless if you see each gets richer phenomenon I don't really mean that they are going to form a separate group. They form a continuum from a few very highly connected notes to the many many many notes but the few connections holy and that's what they did to skip a few times means and really behind this structure that we seem to defend that works appears to be deep. It's got such a phenomena Now this also this points at a potential weakness in the scale free network. If networks were random and we started to take out particular nodes the consequences for the entire network. Would be small but if we have a network the scale free network and we have certain nodes with many many links. Then if we take out one of those one of the and one of the things we called hubs one of the nodes with lots of links that we take out a hub then the consequences to the network are fairy

fairy significant and sort of the model that you can think of as the the way airlines now operate where so many of them connect through major hubs. If a small airport say for example our airport here in Champaign Urbana for some reason it was not operating the consequences to air travel wouldn't be so big but if O'Hare somehow wasn't operating then air travel in the end. Country in the world perhaps might be affected. That's right and let me actually start with better and use not that sort of a defense and for the for the airline ology is a very good one if you will I always suggest the place they want to think about just different or they should think about the airline system. Very indeed down many many tiny airports such as the South Bend Airport which is serving the Notre Dame community here but endowed or connected to a few hops listed like Chicago's O'Hare and you know and that's that their line example also gives too wide the idea fried the hops are important because in a hub dominated network every

time you're on the net travel from one note to another one you have to go through the hub so perhaps they have a sensor Lord role in keeping the network together. But regarding the issue that you just broke talk about their fragility or their sensitivity to attacks. Let me start with the better news there and the better news is that you know we know that many networks are very robust against phasers and what I mean by that. On the Internet at any moment there are a few hundred are out there is a town not functioning to simply just broken down they're going to fix it in a few minutes or if you are so a few days but at the moment they're broken down you have to do or network does not fail a series so you can have several hundred routers not working and yet distilled to one network to be functional to be able to send email send and communicate. There's a similar robustness in our body at any moment that I'm talking to down hundreds or perhaps thousands of all the mistakes in my body. You know their mutations from one. That's another advantage and says more des pli down Mrs. Ford in propane. There are lots of other mistakes going on.

Yet I'm not dropping dead any moment such a mistake happens. Somehow my body ignores and those stores add or what we call it that it's robust against such a loco that Royce. So what we see in general as a general pattern that complex systems particularly network systems are very robust against Randolph a the receiver component found only goes down. Not much will happen. Now what it turns out that indeed the network typology place a very important role. So when we when we investigated how do these scale free networks respond to these random breakdowns where we learned that indeed you can knock out a very high fraction of the Norse from the network and the network will not break apart. This is not true for a random network random network if you knock out the critical affection to know the network would break into tiny islands and don't communicate with each other and ask if you meant to work. You can knock out 80 percent of the nodes and the remaining 20 percent they will be still together in a cluster and they will be able to communicate to each other and that's what we call that

the skiff that rocks have of stability and proper logical robustness or structure offer us number robustness against against random arrows and as you said the reason WHY does not process robust is because they don't have to get there by the hops and even random arrow secure most of the time. He's done many many more nodes they will be affecting the small notes very very you know and randomly knock out the hub because there are so few of them and therefore you know if you knock out the small nobody's going up there for the network and will continue to be functional. But the price of this phenomenal robustness is indeed as you mentioned and the ability of the network to withstand attacks. So if I know you charge the hops and I go and take out the hops one by one on the network we were very easily breaking into PCs and that has very important consequences you know and we know that the Internet is Askia for network here and by I mean computers connected to each other by a router or some physical life. That's right the Internet that I'm talking about and that really means is that there are a few very highly connected

notes or hubs that keep the network together and run of them being your choice Chicago the most connected hub in the United States is in a suburb of Chicago. Now if somebody knows who you are just hops and that's public knowledge you can go ahead and knock five or 10 of them out and the network would break into pieces that cannot communicate with each other. So the Internet rights are very robust against random errors just break down the Soprano's. It's very fragile. Again you've directed attacks if somebody knows what's doing could actually take them out. The same is true for the South you know the South was we when we looked at the how the protein tracked but each other what we found is that you can knock out a very high fraction of the proteins from the cell and not affect its behavior. If you take out some of the most connected hops to sell the rights of a Dyson sound a more selective protein simply in absence of most active proteins of the cells are not survive so what we really see now is that networks have to recall Achilles heel properties stopper a feature that very robust against Randall failer but very franchise

against attacks and this is so certainly bad news for example the fragility is bad news for a for the Internet because indeed to give us an avenue for for hackers to if they want to break down the wall network but it's good news for biology because in biology we don't really have attacks you know like to design attacks against South but we could design docs based on this principle that would attack the hops of some of those That's a bacteria that we didn't by not desiring the human body for example. Oh that's interesting and I guess it is a little comforting if you think about the fact that if you're going to design a system and look at what is the most likely threat that in nature probably random attack random errors are indeed more of a problem than targeted ones and so if you have to take your choice maybe it makes more sense to design a system that will deal well with random errors because then it seems the targeted attack. Rather rather often at least with the kind of networks that we're talking about now.

This is this is going to be there could be a human agent at work so there again you have to design some other sort of sure that there's nobody in the furniture to kind o p prates F against you know against blood to find the pads because any system has when their abilities so I mean any defense could be actually put off. It's really the most of what it makes for any natural system is to somehow evolve into a state which is very robust against inevitably random error. Our guest in this hour of focus 580 Albert Lamar basi He's professor of physics at University of Notre Dame and directs a research there on complex networks and we are talking about that the new science of networks how he and his colleagues there would last few years have really changed the way we think about networks and added to our understanding of how they work. And also part of the idea is here to get people thinking about networks and once you do that. He says that you're going to start to see them everywhere and knowing all of that

and having a better understanding of the way networks work will perhaps give us a better understanding of the way the world where. And if you're interested in reading about this again you can take a look. His book linked the new science of networks. It's published by Perseus publishing is in bookstores now and questions are certainly welcome. 3 3 3 9 4 5 5 toll free 800 to 2 2 9 4 5. We do have a couple of callers coming in to join the conversation so we will start in with someone here in urban on our line number one. Hello. Oh yes can you speak to the issue of redundancy in these complex systems. I recognize the very robust but how much of the function of redundancy. Well this is a wonderful question and there was one of the things that really occupied us at the beginning and live and let me just for those who just may not be familiar with the artist as it would not be familiar to

undersea Probably you mean the fact that there are certain functions that can be carried by different parts of the network and or different links. I mean there are several paths between any pair of notes you know that will contribute to dirty down to see one of the things you may want to realize is that that random networks are just redundant as disk a few networks and probably even more so because the skiffy networks are so much relying on the hops is the sound that he dances taken away from them because of the very important role of the hops. So when it comes to being done to see random networks you know play an act have an ecological redundancy asterisk if networks yet. Random networks are not robust against Randall paper because if you every random that Forth has a valid if I would call a critical threshold if you knock out more noise than the trash or you know like let's say because that's what could be let's say 20 percent if you can knock out randomly more than 20 percent of the nodes the network to break into non communicating islands and that's a very van or desert

for all four from the network fury you have community and for the record the percolation theory community. What is interesting about the skiffle network is that the trash or for the skiff network is the one that is you have to knock out or the North to break the network into pieces. Simply they refused to break apart and so that's really very different reference come seeing that yes if you got into redundancy they have an echo response you know they have acquired it could be perhaps what you done to see when it comes to random was given at 4 but you know Rand on that work. Duran don't break down the notes really break the network into pieces in the scale free network. It does not and this is not only a quantity of asked for it but there actually been proof now in the literature to show that the trash short for escape from that was exactly one which means that you have to knock out all the notes to break the network into pieces. Can we take this one step farther knowing biological systems want to be an ecological systems in particular one of the

assumptions is that provisions to the system some disruption of the suit. Increases diversity does not occur. But you see the same phenomenon occur in things like the Internet. Granted probably not exactly endorsing the same language you see the issue of diversity is really very very specific issue to biological systems and pretty good to ecological systems. And you know I don't want to misled you lots of other reasons why some systems are robust you know there is what we call a dynamic of robustness or what we called our feedback loops in every system that dynamically tried to correct any adverse and Biological systems have very much such a feed feed that looks like in the Saudi competition if a chemical goes very much up another a few other reactions with tar need to shut it down essentially so that it cannot keep the south and. Some kind of dynamic arose to surprising effect perhaps in the new thing over here is that we never expected that the network topology

would also play could play a role so in addition to this dynamic a feedback mechanism that could that contribute to to robustness both in ecological and biological and systems and on the Internet there is a proper logical future. Now I don't have a clear answer to your questions about diversity and one of the reasons is that simple discussion has not been addressed in the terms of thing turn apt and if you want to give some explanation why not is that you have to realize that this with the concept of Giffen effort was introduced to us by 1999 and this is a fairly young young science you know there has been several hundred papers in the literature in the last two years about the subject but very very many questions are just beginning to be asked and one of the areas which has not been asked you know too much coverage into this the church is the ecological systems and religion in fact of ecological network structure and ecological systems and there are several reasons for that. One of them is that ecological maps are very poorly mapped over the run of the best

maps we have a variable in ecological systems of how species are connected out of food vast and reliable food that's contained at most 100 nodes 100 species which is a very small forest deep at another disease. I would say that when it comes to a logical system some of these questions are just beginning to be asked and perhaps in a few years we're going to have asked for an in-depth language that you are asking. Right now I am afraid I don't have one. Thank you very much. Thank you and we'll continue the next caller. He's also in our band and line too. Hello. Yeah I was hoping I could find out more about the way you characterize the Internet. When you when you say that there are these certain how many would you say there are. Yes that's a wonderful question and I guess many many people who work at the Internet would like to know who are the hop stuff that I'm going to have a task for to death and the simple reason ease into town kill free. What disc if reset means really

is that there is no intrinsic skill in the system that would allow me to distinguish the notes from each other saying from here on I'm going to roll the I'm going to call you guys hops and thrown from north that are smaller than descale. Dave will be not tops but we have in the Internet that we have you know if you know that have you know a few hundred links and a few others that have a few hundred links and then we have a larger number that you have like dozens and then even larger that have you know a few and then even larger that have won. So what I'm trying to say that there is a continuum between the most connected very few to the to the very top connected to many many nodes and there is no scientific basis by which we can assign saying from here on you are hob. Before that you know if you're smaller than the threshold you are not the hub and this is very much in contact with the random network concept because in the end all network as we are at full meaning is measured in the beginning of this discussion. They have a better distribution which means that there is an average more than the

system and I could refer everything back to the average No saying if you are much larger than the average then you are perhaps a hobby few are much smaller then you are not. Since we don't have an average node and end in this type of networks I mean average does exist but it doesn't it's not particularly meaningful because of the nature of the distribution. You know it's we cannot scientifically assign trash or saying from here are we called Hobbes. So it really depends on your personal view of what you would claim a spank hops I couldn't or what you would call it bankruptcy all I can tell you I can tell you what these relations allow me to tell you is that that given the size of the network I can tell you how many notes we'll have let's say a thousand links I can tell you how many notes we have a hundred links and I can tell you how many also have a few links so I can have a predictive power to tell you what is really decided distribution I don't have a predictive power saying from Vero you should consider a something as being helpful. Well what about percentile. What would be harmony with many

nodes would there be in the in the top 1 percent power bars. Well OK OK well this is one of the you see what you're talking about just behind us Giffen at crux is there's a popsicle we call a power law distribution so the number of links a note has that follows really a power law and power laws are something which we're curing many other systems and there is this very popular across the court the 80 20 30 percent to 10 year old and essentially what used to eat it and it was tell us in the late language that the system has a power distribution and typically use hey 80 percent of the crime is done by 20 percent the Billings. Oh the very thought of the delinquents that's a present to the population and so on. So there are many examples of that if you took 20 percent to something like that applies. 45 that for example I don't remember the Internet's number 30 member that's on the border drive at about 90 percent of all buildings belong to 10 percent of the sides so 90 percent or so incoming things that the popular tilling silly are all just a bit only along 10

percent a very similar number should be applying to today's Internet and you know if you go to our website and look up some of the papers you will probably find all data that would allow you to find a precise ratio but it should be around 90 10 percent of 2 billion is what you're saying. That tried to get into Word 5 decades yes 10 percent of the 2 billion notes would contain about 10 first. About 80 to 90 percent of all the links that are out there and you're saying that if you targeted for example maybe the 10 percent is wrong maybe it's right you know. But if you targeted the the hub. Yeah you could bring down the system that some point. Well now let's let's let's of course make a distinction between toward the end to internet when I said the 90 10 and 10 percent is that essentially I was referring to the Lord I have described remember the numbers to whatever is a virtual network of five pages connected by you are us. And then there's the Internet which is a physical network. Now if you go ahead and target the Revive No

it's you not going to bring down the Internet you will shut down a website and that's what happened to the Mafia Boy attack. If you target the Internet you know then you could do indeed shut down the network now. How many holes Dorce Internet hops you have to knock out. That's a good question our current maps of the Internet contain about 250000 routers. That's the size of the system we're talking about on the Internet because of the physical infrastructure case and and the question is What does the trash hold and what we found when we did the simulations on this is that in order to make to break the internet into pieces by attacks you you have to knock out a bit less than 5 percent of the nodes that may seem like a lot but that's misleading and I think you can actually achieve the same thing by knocking out much fewer than that. And why is that. Because when you knock out the biggest node Derville be rerouting 0 order. The

packages that are sent to death note to other routers because that node is not functional any more detail does not function automatically. The stuff that actually go through other outlets. If you knock out try with six you know round. Oh if I was six figure out there as they would expect it is really growing to the tiny dollar terrorists abroad but much more outrageous and essentially you're creating a denial service attacks on the smaller one. So if you factoring dynamic a futurist of what's really how the routing is done how the packages are scenting sent you don't need to knock out 5 percent of the know it's on the internet but what's more this but that is the percentage of it that would achieve the same effect. Instead one more question if I could I was wondering if it is the there's a lot of discussion about how the media keeps concentrating into a few networks and I guess there you're talking about. Communication of messages. Is there anything you can say about what about that or is that too small a

system. Does it look to you like it might in the future degenerated into just one hub. Or is there any hope for it staying you know and as diverse as possible that you accept if you get if you get my point. Just just to make it clear to me I have some I mean I figured out how you know what's a second if you mean that the network the media isin after work. Yeah you know with Time Warner Disney Rupert Murdoch Yes complaining about all the concentration. Sure. So and you mean like I censored the globalization and other effects I call that into power and more and more into a few hands a few hands of very thing to say about that or is that too small for him. If you want to concentrate only on the media that's a too small system. But don't cross it in the media to put it on the world economy. You know and closer every company that's bang one node and of different links between them is bank different.

Whatever it's selling buying relationship the business relationship have that we have between them and if you look at that prospect. Then you can try to drawing some conclusions and one of the things I mentioned I'd be precise in answer to your question is that if you have a skiff a network I can tell you depending on the size of the network how many hops we're going to have and how many smaller notes and how many tiny nodes and people who looked at the different aspects of networks in Economy day found the game disk if it apology what this really means is that that there are many many smaller companies and there are fewer and fewer large large companies and there were always a hob to very very connected know that were there. What is interesting is that we expected to see pushes to be certainly stationary but if the network is growing that is for example there is a proposition which opens the media market to not only to be but only not to be interested only in US but they can go to Mexico Canada and Europe or other countries. Then the size of the network has increased and the debt the size of the hops must increase us well because that's what we call a ski

finish and cheap as a metric increases so do the hops. You know we know proportional fashion. So in many ways you can view Deuce Bigalow mortician of power as being a natural consequence of globalization as a consequence of the fact that the network has increased and in order to maintain the skiffy future it has to increase its biggest hops. So that would be like a naturalistic explanation saying you know essentially their economic reason that pork it's a mess it's all about relationship between distinctly different companies and it's a growing network you know because we consume more and more and we bring other countries up to consume more and more and thirty's good you will develop these hops if the network is constrained on a smaller country. The hops will be small or large a territorial reach would let you let the players to operate a larger at the largest notes would be. We just have a couple of minutes left. We have one more. I'd like to try to include. Sure they are also in Urbana.

Yes I believe I heard Professor Barrett basi say that network analysis was a new science. However I think the study of networks goes back to the 30s and 40s. People have been working on our network analysis specifically social network analysis for the last 70 years. Professor Baer basi has done that. That's relatively it was anything but a much larger scale. But the idea of studying networks is quite old and some researchers at places like Harvard and university California Irvine and MIT have some wonderful stuff over the years in network analysis come in for them where the study of structure at these levels is something that perhaps a professor very possibly could. And from I would have been done on a smaller scale. But yeah absolutely I mean I'm sorry if I gave you the wrong impression. Saying that when I give a talk I always tell people hey don't you think that I'm coming here to tell you that I discover networks that by no means has happened. But let me tell you probably what is actually perhaps the new Indeed there has been lots of work in social sciences and if you read laying down I

discuss this work of standing the ground mark on the letters and other sorts of stuff you are just going to Syria and I should mention that one of the most well-known for society network groups is actually and who are not right now. PROFESSOR ROSS some of the groups and he has one of the most classic books into serial social network analysis. But there has been in the last us a very significant change and that came not to be you know a bunch of physicists and mathematicians getting into the game and feeling smarter than anybody else but simply by the availability of data. What has not been available in the 1940s and 50s and not until the last few years is really they found a very very large networks we don't have wards wide Bab to live. That's what the site a sociologist has a chance to study was at most a few hundred notes. What we have learned is that some of the speeches like the skiffy that's what popular GI and many other features are simply not to his approach if you look at the very small network. So in the last year since everything is becoming computerized we started to get the

talent or drive ever looks like how actress I connected in Hollywood to each other because people get Yvette's side school to go to movies you know essentially detail about biological reactions have been connected into half fights before it was scattered or along to the torture now you can find it into databases and disavowed of data has lad has offered the possibility for us and many others to go and look and ask other questions that could have not been asked essentially before it gets much smaller data sets. And and that's what really has created a lot of excitement and indeed the community to a certain degree is a very old associate a sociologist work you know and and we very much incorporate some of their ideas but some of the questions we are asking to disappoint are questions that are quite different from what had been asked before simply because we have to teach a local or two of those questions. I hope this answers your question. And unfortunately we don't really much time left but the sort of

the next interesting set of questions is what what sort of practical application. Do we have for this sort of new. If we do understand networks in the way that we did not before have a better understanding how they work then how do we apply that knowledge. Well damn any leverage. I mean first the first level of complication of course is that you see a new conceptual way of thinking about it and and that often leads to you know practical ideas of how to use them but I should mention one area has already led to a quite different on the stand and it has already an impact on how we look at things and that is the spread of computer and biologic O'Dwyer says you see for many years be dismantled you and many others have actually worked on that area. How receipts are spreading but there is we talk about 8 spreading from one person now devote to sexual intercourse. Are we talking about how a computer virus that spreads from one computer another one and a basic assumption was that if a rider is not particularly wide or land that these you know its very hard to get it to its. Why are you

very hard to pass from one percent to under one then derive despair die out. If riders is to borrow a certain threshold it's a wider lenses about certain threshold States but a spread that is just about all it knows and this transfer process has been very important for beauty. You know essentially stopping epidemics because the reason why we have we are asked to use condoms is partly because potential is going to save the individual who is using the condom from AIDS but we also Florida Chad I mean they've increased I mean decreases the spreading of the age of eight and the fury behind it is that potentially it could die out if enough people use condoms. Now we know the data is not dying god but it's actually spreading at the larger larger rate and to compute to physicists essentially from three guests have offered an explanation about two years ago that had got a lot of attention. What they have shown is that this concept that the riders would die out if it's not the rider and just through the network of the fandom when the skipper network it's

not through on the skiff a network even extremely BQ Veyron flightless is of a spread out reach a very high percentage of the nodes so there is a trash really is really not skiff in Africa zero now. Why is this important. Because most of the networks so much wider seats are spreading our skiffy you know computer viruses are spreading on e-mail network of who we know is who want to email and there has been a study from Germany about a year ago showing that that has the scale apology to spreading to the sexual network of who has sex with whom and there has been just possibly the rest of the Stockholm joint study abroad to show what that show is a scare for nothing. I'm going to my apologies for interrupting you I'm very sorry to have to do that because you haven't quite made your point but we're going to have to stop because we have used our time and first of all I want to say Professor Bob we thank you very much and and also to people who are listening if you'd like to read more on this you can look for this book we've talked about linked is the title by our guest. From Notre Dame. Thank you very

much for talking with us today. My pleasure and did the rest of the story is in the book.