http://kdl.cs.umass.edu/papers/fast-jensen-aaai2006.pdf

The link above leads to an article concerning the most significant network in the NFL: the network between coaches. From the beginning of the NFL to the present, almost every successful NFL coach has been linked to a great coach before him as an assistant under him. This article identifies the network that these links form and explores the reasons behind the development of these bonds.

The article focuses on the success of a coach and how his success affects the development of the coaching network. The main criterion used in evaluating how successful a coach is is the ability to win a championship. The article concluded through this criterion that the fewer mentors a coach has, the more likely he becomes a championship coach and the more likely he produces future head coaches. The article presents two figures that support this theory. On the third page of the article, in Figure 3, a network of coaches is presented. The colored names are coaches who have won either a Super Bowl or a championship. As one can see, most of these coaches has had few mentors, but ended up producing many coaches, some of whom also won championships. Then, on the next page, Table 1 shows a list of coaches who have won 3 or more championships. As one can see, all these coaches eventually produced at least 6 head coaches. This article shows us at the end that the vast coaching network has been developed over the years by coaches of championship and the assistants they mentored.

This article relates to the class in that it shows us the power of a social network. As we can tell from the article, many of the assistants that were linked to successful coaches eventually became coaches themselves. Besides the people that started coaching when the league started, almost every successful coach is at most two degrees of separation away from a former title-winning coach. We learn from this that in the NFL, being part of a strong network of coaches can lead to future jobs and success in the league.

New Social Networking Technology Packs a Wallop

A spinoff company from Microsoft called Wallop, still in invitation-only mode, hopes to redefine online social networking. While the details are not complete, Wallop is a flexible platform that allows people to centralize blog posts, photos, videos, and conversation in one place. More versatile than Facebook, but cleaner in design than MySpace, Wallop aims to bridge the gap between these two sites.

One of the more interesting features, labeled as the “radar,” displays a graph of a user and his/her relationships. Although its meant to be an easy way to communicate with multiple friends at once, it also visually represents the strength of each relationship, as determined by a matrix of possible factors. This has a direct tie to our class discussion of strong vs. weak ties as well as the distribution of power in a relationship.

Wallop, the social network where you pay for pretty

The CNET article above also discusses how Wallop takes on attributes of Second Life, because it allows users to make money. Developers who are skilled with making Wallop widgets and design packages, can offer to sell their products to users - and Wallop makes it easy to market.

Washington Post Article

American Sociological Review Article

The first attached link is to an article from the Washington Post about increasing social isolation in American society. The second attached is a study referenced in the article. The study, conducted using research data collected from 1985, and a similar study conducted in 2004, concluded that the modal number of strong ties has gone from 3 to 0 in the intervening years. (Note: Within the article, strong ties are qualified as social ties in which important personal issues could be discussed freely.) Along with the general decrease in number of strong social ties, the strong ties appear to be bunching together more tightly – more often seen is a blood relation/spouse as strong tie, rather than a friend, or a co-member of a shared group/organization.

As social conditions change and communications technology improves, it would be expected that the number of social opportunities would increase, rather than decrease. Along with improved communication and social opportunities, comes the expectation that stronger friendships would be the result of extended opportunities. However, this appears to be the opposite of the case – twenty-five percent of the respondents in the 2004 survey appear to have no close ties, up from ten percent in 1984.

There are many suggested reasons for this discrepancy, but I will limit the scope of this post to one of the major reasons mentioned in the study. One of the suggested reasons for the difference between the 1985 survey and the 2004 survey is a changing social landscape in American society. Location, recreation, work habits, and most importantly, communication habits have changed, and the network linkages have changed along with them. With computerized social networks and networking becoming commonplace to bridge distances, it begins to change the dynamics of social interaction. Extremely large social networks are created, such as Facebook. However, once you create a very large network, the strength of the ties contained within the network weakens. As described in The Tipping Point (Malcolm Gladwell, p.175-181), social networks eventually reach a critical mass, past which strong social linkages become much less common. Basically, the ability to maintain deep relationships becomes that much more difficult with a large amount of fundamentally different relationships and limited resources to maintain the relationships. The overload means that most of the ties created will be weak social ties, sacrificing local stability, and a “social support base” for a more free flow of information between nodes. This inherent trade-off between network size, structure and information flow is something that should be considered when examining both current social networks, and the dynamics of social networking.

Following what we learned in lecture, the number of traders and the interaction between traders is extremely important in determining how markets are structured. This is evident when a particular market could be structured to grant the trader a monopoly with 100% of the profit of the transaction while another market could be structured with perfect competition and no profit going to traders. Furthermore, we were told to think of traders in the context of stock exchanges. This brings forth questions regarding competition between stock exchanges and related companies and if regulation is needed. In 2002 Cara Schwarz-Schilling and Mark Wahrenburg wrote an interesting paper, Regulating Competition between Stock Exchanges, which in part identifies some of the complexity involved in the competition between stock exchanges.

The major stock exchanges, such as the NASDAQ, have a complex relationship with alternative trading systems (ATS). These include electronic communication networks (ECNs) that work with the stock exchange to execute trades and thereby help to facilitate the exchange business via methods like inhouse systems. Yet, the exchanges use these ATS at a cost and often have a complex relationship with the ECNs because they are competitors taking away trading volume. These ECNs are responsible for a large percentage of trading volume at the major stock exchanges and have grown in number, company size and prominence. One of the most notable ECNs, Instinet, formerly owned and operated by Reuters but now a public company, is a fair example of an agency broker (what we would think of as traders).

The inclusion of these firms is thought to be good for the customer (investor) and the market as a whole because transaction costs are reduced and the market is more efficient (at least in theory anyway) because sophisticated technology links buyers and sellers. However, the market makers at the major exchanges can use this technology to their advantage:

“Market makers use ATS to precommit to pursue a less aggressive pricing strategy in the incumbent market, by committing to buy a part of the overall order flow and committing to execute them at equilibrium prices later derived in the incumbent market. Intuitively, following aggressive pricing strategies in the incumbent market becomes less attractive because lowering the spread ceteris paribus leads to a smaller increase in profits due to a smaller reaction of trading volume as compared to the situation without the inhouse system. The introduction of alternative trading systems may thus have the undesired side effect of decreasing competition at the market-making level.”

As the paper suggests, the customer might also be disadvantaged because there could be less transparency in the cost structure when ATSs are involved:

“The overall cost of trading includes brokerage fees, fees and commissions for settlement and related services and finally the spread paid to market makers. While fees are an obvious and transparent cost of transacting, many investors have only a limited understanding of the amount of money they pay for market-making services in the form of the bid-ask spread. The shift of trading volume away from traditional exchanges towards ATSs may result in a general shift of transaction costs away from transparent items such as brokerage fees towards non-transparent items such as the spread. Future regulation should ensure that customers have access to all necessary information in order to make an informed decision between trading systems.”

To not only aid the investor but bring the market greater efficiency as well, the paper calls for regulation and oversight by government agencies such as the SEC. While regulation issues may touch upon politics that I did not intend to address, it is important to think that possible regulation will relate to the competitiveness of stock exchanges. In essence, the government is structuring how traders interact and how markets are set to match buyers and sellers. Agree with it or not, the markets are being formed and watched over in this manner.

Everyone knows how fast information can spread on the Internet, especially on social networking sites, such as Facebook or MySpace. Everyday new Facebook groups are created not just for fun, but, over the past year, there have been groups forming for specific causes. After Hurricane Katrina, numerous common interest groups popped up, trying to bring people together to help the victims of the disaster. These groups received many members, and although they were just “Facebook groups” they still carried some meaning. There are often times where a disaster occurs, and families of victims constantly try to reach their loved ones, but are unable to, due to inoperable, or clogged telephone lines. What if there was a website that informed people of disasters all around the world, allowed victims of these disasters to reach out for help, and could possibly save lives? Today it is the Internet that keeps the world informed, and the Internet that has become an important medium for all types of communication.

Recently, two researches at the University of Maryland, computer scientist, Ben Shneiderman and expert in human-computer interactions, Jennifer Preece have proposed a new social networking site, 911.gov, which would “allow residents to report disasters, request assistance from neighbors, and check for emergency updates and relief information”. This website would provide up-to-date information on disasters occurring all over the world, and would provide a place where people in these disaster areas could reach out for help from their local communities. The social network would allow for better communication all around during times of crises.

It is interesting to see how online social networks are expanding to not only provide a domain to foster friendships, relationships, and business networking, but may be potentially used to save lives. If this website gains approval, which is questionable due to the cost of developing it, it will be interesting to see how effective the social network will be. I think any website that could provide the public with more information during times of disasters occurring around the world, is bound to be beneficial in some way. I know that during 9/11 and Katrina, finding information about what exactly was going on was difficult, because you could only rely on the media (televised news reports, newspaper articles, etc.), and you only had limited information about the condition of the people involved in the disaster. 911.gov, could provide the much needed extra support for disaster victims.

For more information about the 911.gov proposal click here

http://wow-en.curse-gaming.com/wiki/a-guide-to-dkprpp-explanation-examples-and-summary/

This page is part of a wiki on Curse Gaming, a popular site dedicated to World of Warcraft and other MMOs. It describes market systems for distribution of items in such games when there is competition among many players. When a group of players assemble on a repeated basis to defeat raiding (end-game) content, they want to create a market that allows fair distribution of the limited number of desirable items which will result from killing bosses in the raid dungeon. One way to do this is to assign points to raid members for attendance, then let them spend the points on items as they drop. As the site describes, this can be done via a fixed-point system, where each item is assigned a value which is deducted from the player’s total point value, or an auction. To use terminology used in lecture, two popular systems are an ascending bid auction and a first price sealed bid auction.

Most players decide when an item drops how they value it - this decision is based on the number of points the player currently has (Is the item valuable enough for the player to “blow” all his/her points on if he or she is bid up to their maximum points?), the level of upgrade of the item (How much more powerful will this item make his/her character?), etc.

In a first price sealed bid auction, players communicate their bid to a guild officer or raid leader via a private in-game message, and the item is awarded to the highest bidder. Choosing a valuation for the item in this case requires a player to consider what one’s competition is (If a hunter item drops and there are 3 hunters on the raid, one might bid lower than if there are 5 hunters along that day) as well as a personal value on the item in terms of level of upgrade.

Until very recently, I was a member of a high-end raiding guild in World of Warcraft that used a DKP system with ascending bid auctions. My boyfriend or I ran anywhere from 5-15 auctions on a raid night, announcing in a common chat channel that bidding was open for a specific item, then monitoring the numbers that players typed into the channel until all but one player had dropped out. We then formally announced the winner and the player could loot the item from the boss. We noticed that the valuations of items changed over time - a first kill of a boss with coveted loot could yield prices of upwards of 100 points, while a boss who had been killed many times would drop loot that would go for 10 points or less. (Our guild set 3 points, what one would earn for about half a full raid night, as a minimum bid or “reserve” price. Some systems have a different reserve for different items depending upon their perceived value.) Inflation in the system also affects the price of loot - we had some players for whom the items that were dropping were not upgrades, so they would end up accumulating points and easily outbid other players for coveted items later. (While some players never seemed to have more than 20 or so points, I ended up with a pile of 200+ points quite by accident, and was #2 on the dkp list when I left the guild last week.)

One difference in an ascending bid auction is being able to see who else is bidding and what they are bidding, so fierce competitions can arise. A player’s valuation of the item may change when they see that a particular player or member of a class is bidding - they may choose to “pass” to a friend, or bid higher to beat someone they do not think deserves the item as much as they do. It is likely that players do drop out when the price reaches their value, if they decided on it in advance, which makes the auction equivalent to a second price sealed bid auction as described in class (so the winner pays just above the previous player’s maximum bid).

Example: The Sword of a Thousand Truths drops. I open bidding with 3, in an effort to pay as little as possible. Jim bids 5, Sarah bids 10, and we go up in increments of 5 until Sarah drops out at 50. Jim bids 60, I bid 70, Jim bids 80, I bid 90, and Jim says “out” - he might’ve chosen to go to 100, or hope that 90 was my value, and bid 91 to beat me and minimize his cost. It’s possible I could’ve bid 81 or 85 and beat him, but I don’t know that - so while I beat him in a more efficient number of bids, I also might have overpaid.

Other things to think about: Differing prices of items over time, using a zero-sum acquisition system to minimize inflation, optimal strategies vs. reality (such as setting larger bidding increments to save time).

Warnings Over Privacy of U.S. Health Network

The link above is an article from The New York Times by Robert Pear that raises several important questions about privacy issues in social networks. The Bush administration has a strong interest in creating a large electronic database of private health information that would be shared by different health care providers.

“In 2004, Mr. Bush declared that every American should have a “personal electronic medical record” within 10 years — by 2014. With computerized records, he said, “we can avoid dangerous medical mistakes, reduce costs and improve care.””

Clearly an online medical record would be helpful in improving the country’s health care system as accurate and up-to-date medical information would be instantly available to various health care providers. It would certainly reduce costs as millions of dollars are spent every year on updating and transferring medical information. However, the biggest drawback of this plan is the security and privacy issues that are associated with creating a database of private medical records. As it was mentioned in the article, it is obvious to anyone that such a broad database could easily be misused for harmful purposes. The question then is how does one make this network publicly available to health care providers, yet make it private so that it cannot be accessed from the outside?

This question relates directly to the article that we read during the first week of class, Empirical Analysis of an Evolving Social Network by Gueorgi Kossinets and Duncan J. Watts. The authors studied an email network at a large university and reported their findings in order to analyze their results. However, they could not make all the information available in order to protect the privacy of the students and faculty. Hence, they encountered the same problem; how much information/data can one reveal without severely compromising the privacy of the clients? In this case the data was confidential, but the potential disclosure of information would not have been very harmful. On the other hand, the country’s health care network is a completely different issue. Private medical information in the wrong hands could result in a large scandal that could potentially cost billions of dollars along with the complete loss of medical privacy. A recent investigation into the issue revealed the following:

“Senator Daniel K. Akaka, Democrat of Hawaii, who requested the investigation, said it showed that “the Bush administration is not doing enough to protect the privacy of confidential health information.” As a result, Mr. Akaka said, “more and more companies, health care providers and carriers are moving forward with health information technology without the necessary protections.””

Such a state of affairs could have devastating consequences if health care information becomes available without privacy protection. This is just one of many problems that affects all public social networks. In an age of digital information it becomes increasingly more convenient to record valuable private records in databases that could be accessed by the proper authorities. However, making this information easily accessible also poses many security challenges that directly involve client privacy, and unfortunately there is often no easy way to solve them. Thus it becomes a question of how much privacy is one willing to surrender in return for better health care networks that could save lives.

Bat Batjargal, a Harvard researcher who has written extensively on the social networks of entrepreneurs in different cultures, has some interesting preliminary findings about the role of gender in shaping the structure of entrepreneurs’ social networks. The study concludes:

Men who are entrepreneurs have a lot in common with their female counterparts: They are young, energetic, and focused to the point of being obnoxious, said Batjargal. But they have very different social networks, and use them in different ways. Women have larger social networks for advice and resources. But men, surprisingly, have larger “emotional” networks - the complex of associations that provide warmth, praise, and encouragement. And men apparently profit more from these emotional attachments than women do.

“This is probably the most surprising, counterintuitive finding of this research,” said Batjargal. “Men can be very emotional, and they use these emotional ties better than women do.”

Most interesting here is the concept of different types of social networks and ties; one may have a strong tie with someone on a business level, able to communicate regularly about business problems or plans, but a weak or even nonexistent tie with the same person on an emotional level. A node may serve as a valuable connector in the ‘emotional network,’ holding gatherings and bringing people together to “share the warmth,” but may not play the same role in the ‘resources network’–these gatherings may have little to do with the typical connections associated with business networking. Accordingly, one must consider ties not only by their strength, but also by their type in examining the structure of a network.

On a personal level, the notion of multiple types of networks seems to ring true. Here at Cornell, I have a number of friends that I rely on for emotional associations–going to meals with, joking around, venting frustrations–, but I have an entirely different group of people, other students, faculty, and staff, who I would turn to for advice or for a pointer to a good resource to help me accomplish a task. For the most part, these two groups do not overlap–the upperclassmen I have discussed the Computer Science major with do not know or interact with the friends I eat dinner with–but both groups form an important part of my social network.

Perhaps ever-so-important weak ties can form out of such associations, with one person connecting together previously separated people from an ‘emotional network’ and a ‘resources network.’ For instance, suppose one of my friends from my ‘emotional network’ is complaining to me at dinner that she is frustrated with her course of study and wants to try video game design. I could connect her with a professor running such a program that I know from my ‘resources network.’ The effect is, in the ideal case, a newly formed weak tie between my friend from one network and the professor from another, a tie that may lead to new opportunities for my friend.

(via Ben Casnocha)

Although we’ve seen many illustrations and schematics of networks in class, there hasn’t been much discussion of how we might analyze or model large networks, which are seemingly impossible to understand at a glance. The importance of being able to model large networks rather than simply looking at many examples of large networks is because in most cases merely looking at a visualization of a very large network doesn’t provide much insight into the behavior and structure of the network, as is obvious from looking at the diagrams that attempt to show the Internet as an confusingly entangled mass of connected nodes.

Thus, one approach is to view large networks as random graphs with particular characteristics. There are many papers on this subject, two of which I bring up in this post, mostly because they’re relatively readable for anyone who has any sort of computer science or math background.

The first paper is The Structure and Function of Complex Networks, M. E. J. Newman. Newman gives an introductory description of the role of networks in the real-world and then provides slightly more technical explanations that are useful in analyzing networks of all sorts. The one I found most relevant to this course is the Watts-Strogatz model for small-world networks (Section VI, pp. 27-29). In short, Watts and Strogatz propose a way to model small-world networks that makes them simpler and easier to understand. In general, the Watts-Strogatz model only applies to small-world networks, but in this course, those are just the kinds of networks that we’re studying.

As an additional note, the paper also references the idea of “interlocks” among corporate board members (Section IV.B.4, p. 25), which was the topic of penguin21’s post They Rule, which Professor Kleinberg linked to the paper What do Interlocks do? in the parallel digest blog for this course.

The second paper is Random Graphs with Arbitrary Degree Distributions and Their Applications, M. E. J. Newman, S. H. Strogatz, and D. J. Watts. Although this paper is a little more technical than the one I mentioned above, it’s definitely readable after reading through the previous paper, which gives enough background to understand this paper. The focus of the paper is on random graphs with arbitrary degree distributions, which is in contrast to the small-world network model. In the small-world network model, we make several assumptions about the structure of the network (thus, making it a small-world network), while this paper addresses random graphs where the assumptions are not so constrictive (thus, the “arbitrary degree distributions” part of the paper’s title). As the paper points out, many kinds of networks in the world don’t follow the small-world model, and in other cases, the small-world model misses out on capturing some essential information when used in certain situations and on some types of networks. Therefore, understanding a broader range of random graphs greatly helps in the analysis of many kinds of networks that we find in the real world.

Again, the paper mentions the corporate interlocks (Section V.A, p. 14), which were first mentioned on this blog by penguin21. Keep in mind that there are many other important examples of real-world networks. It’s not that corporate interlocks are the only important example; it’s just that this example is the only one that has been posted on the blog that I’ve been able to connect to the papers (mostly because there aren’t many posts on the blog yet).  Another example the paper mentions that we’ve seen in the class is the relationships among actors and the average degree of separation in the movie industry, which was described in The Tipping Point, Malcolm Gladwell (pp. 46 - 49).

In summary, analyzing networks as random graphs is one of the most effective and practical ways of understanding how networks behave, operate and grow. Most of the analysis requires background knowledge in graph theory and probability. Even those without technical backgrounds can still appreciate the introductory sections to the papers, which outline the significance of understanding networks and their applications, from social networks to information networks to biological networks.

Huge numbers of brain cells may navigate small worlds

            Modeling the brain as a network is just one of the ways in which people have tried to understand the brain.  A “small world” idea was first introduced around forty years ago by Stanley Milgram, who created a study in which participants attempt to send a folder a target individual through other friends and acquaintances.  Of the successful fourth, there was an average of six intermediate nodes (people).  In the late 1990’s, Milgram’s study was followed up by Steven Strogatz (Cornell University) and Duncan Watts (NYU) through simulations involving electric-power grids, actor’s relationships, and ultimately to the 282 brain cells of nematodes (worms).  Neuroscientists have adopted these methods to model the human brain in the small-world setups that are so common elsewhere in the world.  

            In a study by Danielle Bassett, volunteers had sensors at 275 points across their scalp to measure the magnetic field produced by the electrical discharges of neurons.  Six types of brain waves showed up at a different frequency.  From this data the neural network was reconstructed and a “small world” arrangement was seen.  Like so many of the social structures encountered already in this class, the brain obviously turns out to be just like them; there are large clusters of neurons that have a few connections to other clusters.  These interactions between networks act as a unique path for each electrical frequency, and the two highest frequencies during a finger tapping test became synchronized at a certain point during their path.  Supporting the claims made by Mark S. Granovetter in The Strength of Weak Ties, the not-as-strong connections to other clusters help in allowing for the fast transfer of signals between clusters of neurons all the way to the other side of the brain.

            An interesting point brought up in this article is that every brain wave looked similar but just operated on its unique scale.  The fractal-like brain network, when at a high enough frequency, “operates on the edge of chaos.”  According to Walter J. Freeman (University of California, Berkeley), this chaotic behavior is what the brain thrives on.  The article doesn’t go into much more detail about this fractal behavior, but it will be interesting to see more results and conclusions of studies on this.