Original Article

---

eBay is partnering with Bebo to allow users of the social networking site to list auction items on their profiles.The agreement will allow Bebo’s 30m members to detail items they want to buy or sell. Clicking the link will take visitors directly to eBay’s own pages, the Wall Street Journal said.eBay hatched the plan to attract a younger audience. The auction site has a higher than average number of users above 35 years of ageat 58%, while 54% Bebo’s user base is under 18, a far higher figure than even rival MySpace.eBay previously partnered with Facebook to present students with listings for textbooks.

---

World wide web has allowed individuals to at first share information via web pages. In the beginning stages this “information” was only technical and I am pretty sure that nobody would have expected the internet to develop in the ways that it is being developed now. The information is still in web pages, but in oneweb page one can sell/buy products, network with individuals, send/receive e-mails and subscribe to news feeds to be on top of the current world. The two of the most well-known websites for college students are the sites of Ebay and Facebook. With ebay’s strategy to target younger audiences, this move is sure to advertise the name of “ebay” and it will also facilitate buying/selling items between people who might not have been that into ebay.Ebay’s auction type is that of “second-price sealed-bid auction”: however users are given the id of the people who are bidding. Social network websites such as facebook and bebo have greatly facilitated the understanding of social networks because it helps in demonstrating many concepts discussed in class. For instance, triadic closure is made extremely easy with facebook: since the person’s friends are viewed over the profile page, one can easily make friends with his/her friend’s friend. Moreover, friends can be organized in terms of friendship, class, school, region, and many other different categories therefore it can facilitate the understanding of the differences between networks organized into different categories. Then ultimately what does ebay’s collaboration mean to us? You never really hear about triadic closure between items. Items making friends with other items? ebay lists all the items when you search for them, however you cannot really see any relavance between closely related items. there are links such as “you might also be interested in:” however one seller might be from the United States and one seller might be from Hong Kong, and this creates dillema for many buyers. With the collaboration of a social network and an auction function, this dillema disappears and closely related items (not necessarily in terms of the item itself, but in terms of the seller) and other network-aspect ideas can be applied between items now. And one of ebay’s ultimate problems - bad sellers - can be prevented because you actually get to know the seller network-wise as you purchase items.

http://www.leaonline.com/doi/pdf/10.1207/S15327663JCP1303_07

In class, we have studied the fundamentals of game theory and several real-world situations which have been analyzed and/or predicted with game theory. In the article, Failing to Suspect Collusion in Price-Matching Guarantees: Consumer Limitations, Chatterjee, Heath, and Basuroy describe their experiment where they investigated how effectively consumers recognized collusive situations in retail environments.

Price Match Guarantees (PMGs) are the devices which game theory predicts to be collusive. A PMG is a store policy which entitles a customer to a refund of the difference between the store’s asking price and a competitor’s price.

The article describes multiple market situations involving PMGs.

1)      In a hypothetical market – say two electronic retailers, store A and store B – where neither store offers a PMG, they will each compete on the basis of prices and, due to competition, the prices will reach a low Nash equilibrium.

2)      In another situation involving the same market, one of the stores offers a PMG and the other does not. Now if the PMG store lowers its prices, the non-PMG store will also lower its prices. However, if the non-PMG store lowers prices, the PMG store need not lower its prices because consumers will still have the lower price available to them through the PMG. This allows the PMG store to maximize its profit from customers who are not as price-sensitive while keeping price-sensitive customers.

3)      In the last situation, both stores institute PMG. As stated in #2, a PMG store loses its incentive to lower prices to compete. In the case that both stores involved have PMGs, the stores need not worry about the other store lowering prices, so they do not lower their own prices either. The article sees this situation to mimic collusive pricing behavior.

The experiment that the article describes examines whether consumers interpret a PMG as an indication of low prices and whether they detect a collusive situation. They found that many consumers did indeed lower their price sensitivity in response to a PMG and most consumers did not detect collusive pricing.

These findings suggest that government consumer protection with regard to Price Match Guarantees may be necessary. The two largest consumer electronics retailers in the country, Best Buy and Circuit City, both have a type of Price Match Guarantees – a price-beat guarantee – and these sort of policies seem to be cropping up all over the place. Next time you’re making a purchase from a store offering a PMG, buyer beware, that policy which guarantees low prices may be responsible for raising prices!

Link

This article talks about how “network theory” can be used to find terrorists. People studying networks have found “surprising commonalities” between airline routes, people’s interactions at cocktail parties, and crickets’ synchronization of their chirps. These people study networks to have an easier time figuring out non-trivial patterns. New algorithms are being developed and used in exposing these patterns in large chunks of data, like examining e-mail traffic of 43,000 people at
Columbia by Duncan Watts. Stanley Milgram found that most pairs of Americans can be connected by six different people, called the “degrees of separation”. In another mapping, Valdis Krebs decided to map the 9/11 hijackers by starting with two of the plotters and produced a chart of the interconnections within the group. All of the 19 hijackers were tied to one another by just a few links, while many of them converged on the leader, Mohamed Atta. Interestingly, before 9/11 an Army project attempted to map Al Qaeda by “identifying linkages and patterns in large volumes of data,” which may have succeeded in identifying Atta – but he was among a lot of other possible suspects. The problem is that most people are connected to hundreds of thousands of people by only three degrees of separation – leading to many false positives. In analyzing these gigantic networks, some look for network hubs – like Google or Newark airport that have many more links than the average node. However, it cannot be easily done. Similarly the “strength of weak ties” in network theory can be also the key to analyzing these networks. Even if the hubs can be identifying in the terrorist groups, the story is not complete. Like Watts said, “If you shoot the C.E.O., they’ll hire another one.”

This article mentioned a lot of things that we learned in networks. A lot of phenomena are mapped by making people “nodes” while the “edges” represent the relationships and interactions between the nodes. It mentioned the Duncan Watts paper, the “degrees of separation” from the Tipping Point, and the “strength of weak ties.” Identifying a terrorist network is not as simple as analyzing the e-mail traffic at Columbia University. The answers could be in identifying hubs or in the “weak ties.” With so much volume of information coming in, even if it was possible to reliably identify the hubs – these networks are so large and robust that eliminating the hubs would hardly affect them. In addition, it is harder to create a network of terrorists than a network of college students who send e-mail, because the members of university are known while the terrorists are not. While the patterns of terrorists are only beginning to be figured out, these network theory concepts should be helpful in creating a map of the terrorists’ network.

So far we have discussed very basic mathematical methods in analyzing social networks, using basic graph theory and other mathematics traditionally used in network study. This paper* presents an analysis of social networks using a strictly physical analogy—that is, applying laws of kinematics within a closed set of people to show that clear social networks emerge. What is surprising, however, is that the model that comes out describes characteristics of normal social networks that we see all the time.

The model uses particles to represent individuals in a network, and collisions between particles to measure interactions between individuals. Particles react to collisions by changing speed and direction which then could result in more collisions, just as interactions between people change according to previous interactions. Surprisingly, however, is that given a constant number of particles, the system will eventually reach a “quasistationary state,” a sort of equilibrium that approximates many features of social networks. The authors show that in this state, properties such as shortest path length, clustering coefficient (how clustered some groups are) and degree distribution are similar to a social network measured by conventional means (the authors took data about friendships in comparison). Communities arise just as in social networks via this model.

The authors extended this model for specific applications: for example, the spread of HIV could be real-life circumstances just by adding indication of sexual contact to the model. The implications of this model are that statistically, random collisions, if described accurately, are similar to the random social “collisions” present within a social network. While particle motion does not exactly model human interactions, given a statistically significant sample, similar properties do emerge. This model thus gives a natural instance of human social networks, one that does not involve polling a large group of people to get results. If this model can be applied to different circumstances, social network analysis may become simpler to achieve since empirical data can be found without polling large samples of people.

*Gonzalez, Marta C., Lind, Pedro G. and Herrmann, Hans J. System of Mobile Agents to Model Social Networks. Physical Review Letters. 96, 088702 (2006).

http://www-personal.umich.edu/~mejn/papers/mpnsb.pdf 

In November 2002, a respiratory disease occurred in
China, and within eight months spread to over 25 countries, taking over 700 lives.  The cause of the disease, which came to be known as the Severe Acute Respiratory Syndrome (SARS), was a never before seen coronavirus, and its rapid spread became the subject of study in the field of network epidemiology.  This paper reveals that traditional models of disease spread were inadequate in reproducing the actual spread of the virus, and presents possible explanations.  It also presents a more accurate approach to modeling the spread of infectious illnesses.

In traditional models of disease spread, a quantity called the basic reproductive number, R₀, is defined as the average number of people to whom an infected person spreads the illness.  One can easily see that according to this theory of spread, if R₀>1, an epidemic results.  Estimates of R₀ in the case of SARS were between 2.2 and 3.6; this should have resulted in anywhere between 30,000 and 10million cases of SARS within the first 120 days.  However, observed incidence of the illness was much lower, and the authors present possible reasons for this.  Briefly, though, all members of a population are not equally likely to be infected, or to infect others.  For instance, many transmissions took place in crowded places, and particularly in hospitals.  A small portion of the population, then, runs a higher likelihood of being infected (or infecting others), while the majority are at considerably lower risk.

The authors also highlight the importance of the ‘initial conditions’.  The first infected person in a population is called ‘patient zero’; treating ‘patient zero’ as a node in the network, the initial condition is the degree of ‘patient zero’, which we recognize from class as the number of edges leaving this node.  Figure 4A in the paper illustrates how the risk of an epidemic increases sharply with the degree of ‘patient zero’.  The authors compare the spread of SARS in Toronto and
Vancouver.  In Toronto, ‘patient zero’ was the matriarch of a large family who died at home, whereas ‘patient zero’ in
Vancouver returned to an empty home and was hospitalized soon after his return.  More than 200 cases appeared in Toronto, but only 3 in
Vancouver, and these seemed to have been imported. 

This pattern is not surprising; the behavior of networks is highly nonlinear, and it is easy to see that the spread of a disease like SARS would vary exponentially with the degree of nodes in the network.  The point, though, is that the topology of the network is of critical importance in determining the spread of illnesses, and needs to be modeled accurately.  Quantities like R₀ are useful only to an extent, and traditional models do not adequately take network topology into account.  I have left out the details of the analysis presented in the paper, but briefly, the authors examine three (topologically) different networks with the same values of R₀ and demonstrate that the networks exhibit very different probabilities of an epidemic occurring.  In particular, the occurrence of individuals who are labelled ‘superspreaders’ and ‘supershedders’ has a tremendous impact on what happens.

(If you’ve got this far, you might want to take a look at the paper.  It’s not that long, surprisingly readable, and actually quite interesting.)

Some of my colleagues posted earlier about the applications of network analysis in the context of counter-terrorism and national security. BigT referenced a fascinating article by network analyst Valdis Krebs, Uncloaking Terrorist Networks , in which Krebs used public domain knowledge available shortly after the September 11th attacks to identify Mohamed Atta as a key figure in the plot, based solely on social network topology metrics.
As an addendum to this extraordinarily relevant example of network analysis, there has been a point of controversy regarding a DoD (Department of Defense) task force known as Able Danger, which was formed to identify Al Qaeda operatives through similar social network analysis techniques. Able Danger allegedly actually succeeded in uncovering Mohamed Atta as an Al Qaeda operative a year before the September 11th attacks, so there has copious dispute regarding why their findings were not acted upon.

In the same vein, there certainly have been a large number of high-profile stories involving the application of network analysis toward national security problems, especially with the many recent controversial DoD and NSA programs (e.g. Total Information Awareness, warrantless wiretapping). However, basic network analysis techniques were developed out of necessity and used long before the recent formalization and recognition of the field. Intelligence gathering is an intrinsically network-based field. Human intelligence (HUMIT) relies on the cultivation of a network of assets, whereas financial intelligence (FININT) is comprised of the investigation of transaction networks and other monetary connections. In the paper, Terrorists/Liberators: Researching and dealing
with adversary social networks, Karl van Meter explores a variety of historical examples of different types of network analysis accomplished by various intelligence organizations. Though while van Meter examines these intelligence gathering techniques from a historical context, their relevance and continued efficacy in the dynamic world of should not be ignored. I will present a few techniques that van Meter discussed, contextualizing in terms of graph theory.

Social Network Analysis – Nodes are people and edges are their contacts and relationships. The innate usefulness of . The example van Meter presents concerning the first major application of this technique in intelligence gathering is the McGehee’s villiage method. In the mid-1960s a CIA operative named Robert McGehee was sent to Thailand to gather intel on the spread of communist influence in the region. Dissatisfied with the inaccurate data gathered thus far, McGehee had his team engage in an anthropological-style survey of entire villages to flesh out an area’s social network. An analysis of the network’s structure and anomalies led to great success in discovering communist party members and weapons dealers in the region. However, McGehee later became a vocal critic of the CIA. This later criticism partly stemmed from his feelings that many of his findings were suppressed as they painted too bleak a picture of the extent of communist support in the Southeast Asian region (and would have been an indicator as to the futility of military efforts in Vietnam).

Traffic Analysis – Nodes are people and edges are instances of communication (e.g. Phone calls, letters, email, wired monetary transaction). Traffic analysis . Often communications are encrypted and cryptanalysis resources can not be devoted to decipher them all. However, there is information encoded in the structure of the communications network that can be of immense value in identifying members of an organization or key individuals within an organization. Traffic analysis was developed in its modern form by the British internal security service MI5. Though, ironically, one particularly intriguing example of traffic analysis occurred when the IRA adopted similar tactics against anti-IRA operatives in Northern Ireland. Through traffic analysis the IRA was able to discover that all MI5 agents and informants were paid on the same day, and by staking out certain ATMs in the region, the IRA was able to uncover many of the MI5 assets in Northern Ireland.

Movement Analysis – Related to traffic analysis, this mode of analysis involves compiling information regarding a group of individual’s whereabouts and transportation activities during a period of time. Through various statistical and network analysis techniques, the data from movement analysis can yield information regarding the structure of an organization, its key members, and its more peripheral members.

http://www.post-gazette.com/pg/06363/749317-96.stm

This article, published Dec 29 2006 in the Wall Street Journal, discusses a new approach to networking in health care. In the past years people have started searching the internet more often as the first source of medical information. In light of this, the medical community has started to take advantage. Both the CDC and American Cancer Society have experimented with online virtual communities, and both have participated in events in Second Life. In addition, various medical organizations are supplementing the normally dry statistical data on their sites with community forums where people can discuss opinions and share information about the relevant topic.

One of the more interesting aspects of the article is the topic of support networks. For a long time doctors have supported patients getting into support groups, but for patients with rare diseases or patients that live in isolated areas, it may be difficult to find this resource. It’s interesting to see how this new outlet with professional guidance will benefit patients’ well-being. This article combines a few of the types of networks we have discussed in class. It is simultaneously a social network and an information network. As one patient is quoted as saying in the article, “we are allowed to have fun and even joke about things”. An interesting variation of this theme could be one in which only patients are allowed to post, but the patients’ doctors are allowed to read the posts. In this way, doctors can determine the effects of treatment on the patients as described in a social and not medical setting.

Networks such as the fluwikie provide a forum for discussion that equalizes the information flow. Although as before, the power resides mostly in the hosting organization, information can be shared amongst the consumers of health information. For example, since the CDC cannot optimize implementation of pandemic plans in all situations, allowing people from similar communities but different geographical locations to discuss their implementations will allow for the most efficient sharing of information. The host can even redistribute power amongst the users if certain users provide valuable information. This can be done by having time restrictions on posts or assigning weights to certain posts, much like google does with websites.

The Shi, Adamic & Strauss paper “Networks of Strong Ties” (pdf, arXiv) analyzes the importance of strong ties in a social network, in contrast to “The Strength of Weak Ties”. If we recall, the strength of weak ties resides in their ability to connect nodes that are not normally in close contact with each other. The benefit is that more novel information can flow between nodes that are not part of one strongly connected component because the nodes in this component are likely to all have similar information due to their close contact. In “Networks of Strong Ties,” the authors argue that while weak ties are useful for spreading new information, obtaining jobs, etc., strong ties play an important role when trust is required in the dissemination of information, and particularly to form the stable foundation of a social network.

Most of this paper was devoted to determining the importance of strong ties empirically. The authors performed simulations using data from two online communities, Club Nexus which comprised a good portion of the student population at Stanford, and the network of AIM links created by the website buddyzoo.com. They did this by altering the data in various ways and analyzing the consequences of these modifications for the network as a whole. They first defined a “threshold 1″ weak tie as a connection between A and B that does not close any triads (implying A and B have no mutual friends) and a strong tie as a connection between C and D that closes at least one triad (implying C and D have a mutual friend). By analyzing the buddyzoo.com network, they found that removing all weak ties had a fairly negligible effect on the size of the giant component, as it dropped from containing 88.9% of the community to having 87.5% of the community. In addition, they analyzed the average shortest path between any two nodes, and this length increased from 7.1 to 7.3 hops after removing the weak ties. I found these small changes surprising because I would have guessed that weak ties greatly shorten average path lengths due to their ability to span strongly connected components. The researchers experimented with tie strength thresholds other than 1, and were able to obtain a smooth relationship between tie threshold, giant component size, and average shortest path (see graph on pg. 4). Their findings led them to conclude that strong ties in this particular network were very robust and that the connectivity of the network could not be compromised by just removing weak ties. Using graph theory, they also proved that the strong ties in a randomly generated graph are not nearly as robust as the strong ties in the social network they studied. This evidence indicates that there are intrinsic qualities in social networks that are measurably human. I found this interesting and wondered about an algorithm that could distinguish a human social network from an artificial one using statistics such as the strength of the strong ties. If this hypothetical algorithm were able to determine the probability that a given large graph represents a real face-to-face network, what would it say about various virtual social networks? Are some types of online communities more similar to real-life social networks than others? What properties of virtual networks differ from real-life networks? Do strong and weak ties play different roles in the real world and the virtual world? I find these questions very interesting, particularly as we find ourselves spending increasingly more time on virtual networks without fully understanding the ways in which they are changing the way we interact with each other.

http://www.chron.com/disp/story.mpl/ap/fn/4584489.html

One of the largest and well-known online dating services Match.com (www.Match.com), has recently acquired two overseas online dating services, successfully increasing its social network by millions of people. According to an article in the Houston Chronicle, “Match.com to Announce Overseas Expansion,” the French online dating service known as Netclub and the Chinese online dating service eDodo have recently joined the Match.com network.  Members of Match.com will now have the possibility of creating ties with even more people from around the world.  With the addition of these two networks, Match.com will expand its network by approximately four million users in France and 180,000 users in China. 

An interesting detail to take note of is that with the acquisition of eDodo, Match.com will have its first and only connection to China.  This newly created edge to China could be what we referred to in class as a “local bridge.”  It is a bit unlikely that the removal of this newly created tie to China would completely disconnect the nodes in China from the users of the Match.com network, resulting in two graphs that are completely isolated from one another.  Severing this tie would just significantly increase the distance between nodes of the Match.com network and the users in China. To connect a node from the Match.com network to a node in China, the path would encompass nodes of other countries before making the connection, in lieu of a path that travels over one direct edge connecting the two graphs.

Not only does Match.com have its first tie to China, it now has in its possession the second and third largest online dating networks in France.  Over time, it can only be expected that Match.com will expand its reach even further, creating ties between more and more people from around the globe.  With every new conquest Match.com makes, it will become one step closer to securing its foothold as the largest online-dating service in the world, a prime example of the idea that out of all the networks, there can only exist one “giant” network.  (According to comScore Media Metrix, Match.com is already the world’s largest online-dating service.)

 

Article 1: http://www.usatoday.com/news/washington/2006-02-05-nsa-telecoms_x.htm

Article 2: http://arstechnica.com/news.ars/post/20070125-8701.html

The two articles above discuss the power behind social networks. The first article, Telecoms let NSA spy on calls, talks about the NSA wiretapping procedure and the legal implications of this procedure. Each telecommunication company is in control of a large network of users, or customers, which can be depicted using a graph. If all the users and the company itself are nodes, and edges represent a connection between the telephone service provider and the customer, then the resulting graph exhibits the power of the telecommunication company. When the NSA gained control of the Telcoms’ records, they obtained a massive call database which would then need to be processed and searched. (The establishment of the connection between the NSA and the Telcom is a good example of a bridge since it is the only way for the NSA to learn about the telephone network.) In order to search though all this information, the NSA focused their efforts and primarily screened international calls in the hope to learn about unknown terrorist plots and to find US connections to these possible plots. This example shows how nodes with power have great responsibility over their actions. In this particular case, the Telcoms have power because they have a high betweenness and dependence compared to the other nodes. When the Telcoms cooperated with the NSA, multiple legal implications arose because of the contractual relationship between Telcom and customer. It is important for powerful nodes to respect their connections (by not taking advantage of the customer’s dependence or the Telcom’s ability to exclude) since each edge represents a greater profit for the company.

The other article, similar in nature although highly opinionated, entitled, CIA uses Facebook, NSA wants social networking data, also talks about the power of social networks such as Facebook, MySpace and how the government may start to look to these social networking sites to stay a step ahead of crime or as the article says, “keep money from flowing into the hands of terror groups, mobsters, and other unsavory characters who need to launder their cash.” The article concludes that chances are there are no ulterior motives behind today’s popular social network sites but it is clear that the power these site yield should be respected.