Learning about auctions in class recently was an…interesting experience.  To tell the truth, I had very little knowledge on the subject beforehand, so my mental image of the process consisted solely of men in suits waving their little IDing numbers as someone raises the price.  Therefore, it came as quite a surprise that what I now know to be called Ascending Bid, or English Auctions, was actually just one in a wide variety of auction types.  From minor variations such as the Dutch auction, where prices go down until someone buys, to seemingly strange ones like the Vickrey auction, where the winner pays the second highest bid, to ones that seem flat out unfair, like the All Pay Auction that has every bidder paying their bid, but only the winner getting the prize.  What seemed even more amazing was that most of these auctions have an optimal bidding strategy.  With a bit of knowledge, you can figure out the bidding price that maximizes your profits.  However, after a while the initial interest started to wear off.  Even with the variety of auctions, if everything were so cut and dry as to simply bid the optimal price, what difference is there between going to an auction and simply shopping at a supermarket?

After looking around the web in an attempt to renew my interest, I found a variety of tasty tidbits.  Ironically logical advice, such as the advantages of attending auctions in inclement weather (less competitors), and practical tips on what to sell and what not to sell, kept me entertained for a good bit (I am easily amused).  One of the things that most attracted my attention, though, was the concept of Auction Fever and its ilk, discussed in a question and answer article with Harvard professor Deepak Malhotra.  The basic concept of Auction Fever is pretty simple:  the competition, the heat of the moment, and the awareness of others watching making people bid far more than they normally would.  A related topic, Escalation of Commitment, brings up the interesting point that the effort gone into trying to obtain the item can be counted as a negative payoff should the buyer not manage to win.  The third concept of the trio, the paradoxical Winner’s Curse, deals with how those who don’t have a clear idea on the true price of an item often overpay if they win, resulting in a miserable state despite their apparent success.  It goes on to suggest ways for businesses to use these principles to their advantage, with auctioneers deliberately using competitive language to putting the figurative spotlight on bidders to increase audience awareness.  With these additional factors thrown in, watching an auction progress could become decidedly more interesting.  Humans are not, after all, always rational beings.  There is even evidence that those experienced in going to many auctions does not decrease in irrational behavior.  Therefore, a little competition, a bit of clueless ness, or a slight change in setting, almost guarantees to change an orderly little auction into something a lot more exciting.

Read about Auction Fever and the like at:  http://hbswk.hbs.edu/item/4661.html

djr45 discusses a New York Times article about congestion Kennedy Airport, in “Auctions: A Potential Solution to the Growing Airport Congestion Epidemic.” Auctions can be used to limit the number of take offs and landings and as result encourage airlines to use less congested airports or fly during off-peak times. The auction solution proposed could also be used to help solved the problem of which plane gets to take off or land ahead of others.

As airport delays and congestion increase it becomes increasingly attractive for executives, corporations and wealthy people to use private planes. At this time the Federal Aviation Administration (F.A.A.) employs a first come first serve policy when assigning runway clearance for take-offs and landings.  This means that a full commercial flight may have to wait on the tarmac while other small private planes take off.   The first come first serve policy makes private flight especially convenient for busy executives and wealthy people. The number of private flights continues to grow and they are beginning to cause congestion in the sky.  As Kennedy International Airport found that there is finite amount of take offs and landings in a given day so too is there a finite amount of space in the sky.  Although there are ways to encourage commercial flights to use alternate airports, private flights often fly to different airports than commercial planes, these small planes still take up space in the sky and it falls on air traffic control to ensure that’s these planes do not collide mid-flight. On any given day the Air Transport Association cites that between 20 to 30 percent of traffic above New York are corporate flights.  There a discrepancy between the congestion that private flight causes and their financial contribution to the air traffic control system; the F.A.A. estimates that private jets account for 16 percent of air traffic, but contribute only 3 percent to the air traffic control system.  To help ease to burden on commercial flights to support the system a Senate Commerce Committee bill last August proposed that every takeoff should be charged 25 dollars.

A second-bid auction for take-off slots could also be used to decrease the present economic attractiveness of private flights.  With an auction, maximum social welfare could be reached between commercial and private flights.  If auctions are used instead of a flat fee the Senate Committee had proposed, market-clearing prices can be achieved based on time of day and airport such that high occupancy flights may have more preference in choosing their takeoff and landing schedules.  An outcome might be that people who value on-time arrivals will fly higher bidding flights which to leave on-time and people who are not on a tight schedule can fly budget flights that bid lower.

References:

http://www.nytimes.com/2007/08/26/weekinreview/26schwartz.html?emc=eta1

http://edition.cnn.com/2007/TRAVEL/business.travel/10/18/executive.pilots/index.html

http://blog.washingtonpost.com/travellog/2007/08/paying_a_fair_share_1.html

A giant component is a connected set of nodes that consists of the majority of the network. In many networks, the formation of a giant component is unavoidable. I investigated the time taken for a giant component to develop in a social network. I varied the rate at which people made friends, and observed the effect this had on the development of the giant component.To facilitate my analysis, I first built a virtual world containing 1000 people. Each person was immortal, non-reproducing, and had no initial friends. Every day people made and strengthened friendships according to the following precedence:

• Completing triadic closures
• Promoting weak ties to strong ties
• Forming weak ties with a stranger

I varied the rate at which people made friends, and ran the simulation until a giant component formed. I recorded the the time taken for the component to form, and the corresponding rate of friendship. The following two graphs summarize the simulation summary:

The graphs above contain data from the simulation I described earlier (blue dots) and a new simulation (green dots). The new simulation is just like the first, but with a population that dies and asexually reproduces. I made this modification to make the simulation more realistic. However, as seen in the graphs, this modification had no noticeable effect. This is probably because the giant component forms so fast, that an insignificant number of people get to die and reproduce during the simulation run.

Conclusion:

If L = time taken for giant component to form, and T = the average time taken for a person to make a friend : L is directly proportional to T. For this experiment, the constant of proportionality was 2.5 :

L = 2.5 T

My simulation source code can be found here (Visual C++ 2005).

Just as networks constrain how buyers and sellers interact, they also constrain how traders interact in the world of mutual funds. Retail investors often trust companies such as Fidelity Investments to manage their money and make trades on their behalf. Clients’ returns reflect, even if only slightly, the Fidelity traders’ performance, which require them to interact with outside brokers. Behind traders and brokers, however, are humans, and their trading decisions may be affected by many factors. This situation is analogous to a network with varying degrees of edge intensities.

For example, if a client wants to purchase some shares of a company, his trader will locate brokers with clients looking to sell those shares. While the trader’s goal should be to secure the lowest price for his client, this does not always occur. Sometimes, brokers who have ingratiated themselves with the trader may end up with the sale, even if at a slightly higher cost.

The US Securities and Exchange Commission (SEC) alleged Fidelity traders have done just that, as published in an article in today’s Boston Globe. The article explains how the SEC accused Fidelity stock traders of accepting millions in gifts from brokers. Fidelity settled the case, agreeing to an $8 million fine. If these charges are true, it would mean that traders would be less likely to seek out the best deals for their customers, choosing instead to rely on potentially higher cost deals from these gift-toting brokers.

While the SEC allocated significant resources to investigate Fidelity’s activities over several years, this incident shows that similar examples may go unnoticed on a daily basis on a smaller scale. This example shows that many factors appear in actual markets that are not considered in ideal networks such as those reviewed in class. While personal contacts often play a large part in business networks, one must always consider the legal and ethical limitations so as not to cross necessary boundaries.

http://www.jstor.org/view/0361915x/di010136/01p0124k/0 While reading some of the posts written over the past couple of weeks, I came across one about the relationships between the Cold War arms race and game theory. After taking a class last year on British Foreign policy, I remembered a different kind of war that the game theory we discussed in class could be applied too. Sure enough, after a quick search on JSTOR, I tracked down an article that was pretty much exactly what I was looking for.While it is rather old, Levhari and Mirman’s paper does a good job of applying the Nash guidelines to a then-current issue. In the article, the authors build a model based on the international fishing market between Iceland and the United Kingdom in the late 70’s. In this game, the two nations are trying to maximize their own ‘discounted sum of utilities’; simply defined, this is the profit of each country considering all externalities (i.e. the change in market price caused by their opponent’s catch and the future fish population). Furthermore, each player is fully aware of the other’s actions, and reacts accordingly. Surprisingly, in the competitive solution solved by the authors the fish population has a tendency to grow towards infinity, while the Nash Equilibrium entails an increase in consumption by both players, forcing the stock closer to extinction. These solutions are influential to both the U.K. and Iceland on a social and economic level. Rather than encouraging more competition between the nations for fishing rights in international waters, the paper proves that collusion between the two players will maximize profit for both parties. Subsequently, their paper sets the groundwork for future analysis of natural resource disputes; while the conflict is indeed simplified, the model created can be used to help ease tension and even solve other international conflicts.   

Our discussion of search engines today really piqued my interest. I was curious about how information and social networks might be related and read about “social search” which is being pioneered by companies like Delver http://delver.com/, a “ semantic social graph search engine”. Delver is principally interested in improving search quality by better understanding user demographics (age, gender, etc.) and by leveraging those users’ social networks to find content that is uniquely relevant. Delver is betting that users want to be able to search for information that has been generated by their social graph (i.e. blog posts, photos, videos, etc.).

I have observed another use of the social networks for search that might be useful for Delver and other search engines. Users have started to appropriate their status messages as search boxes, leaving queries to be interpreted by the members of their network. One of my friends recently changed her status to: “making hummus… what is that last ingredient?”. I asked her how successful her search had been and she said she had received four responses in the first ten minutes after she had posted. I assumed that the respondents would have been close friends with strong ties, but my friend was quick to point out that unexpected people had responded - the most memorable was a friend of her younger sister. We learned that weak ties were especially helpful in finding jobs, and perhaps weak ties are also valuable for different kinds of queries (the answer was tahini and all four got it right).

Delver Related Links:

http://www.techcrunch.com/2008/01/28/delver-comes-out-of-stealth-with-a-new-twist-on-social-search/

Interestingly, Google also recently released their Social Graph API but instead of improving search, it is more focused on consolidating all of the social networks to build some kind comprehensive social graph, or “a global mapping of everybody and how they’re related” especially for the development of new socially motivated services. “It’s recognized that users don’t always want to auto-sync their social networks. People use different sites in different ways, and a “friend” on one site has a very different meaning of a “friend” on another. The goal is to just provide sites and users the raw data, and they can use it to implement whatever policies they want.”

Perhaps Google will expand its focus into leveraging the social network to improve search quality (relevance, recency, etc.) in the future. It might help them to overcome some of the challenges we spoke of today in lecture related to information retrieval, and understanding (or at least better predicting) user intent. I also wonder if the kind of hubs discussed today in lecture in the context of information networks might also apply to social networks.

Social Graph Related Links:

http://code.google.com/apis/socialgraph/

http://bradfitz.com/social-graph-problem/

http://googlesystem.blogspot.com/2008/02/social-graph-api.html

http://radar.oreilly.com/archives/2008/02/google_social_graph_api.html

In a simplified view, the public Internet can be viewed as a large interconnected network.  If Node A in the network is able to connect to Node B and vica versa there must exist a path between the two nodes.  This path undoubtedly passes through other nodes in the network (unless Nodes A and B are directly connected).  This path is virtual and may span over countries, continents, ISPs and even satellites.

Information between Node A and Node B are divided into packets that traverse the nodes between the origin and destination.  The packets contain information such as Host Address, Destination Address as well as the actual data.  When Node B receives the packet from A, the Host Address is still intact and therefore Node B is able to determine where the packet originated from.  Therefore, any website (Node) that you visit (send data to) on the internet is able to determine the source of the data (packet), you (Node.  Let us ignore NAT and other 1:N types of routing)

Tor allows Node A to anonymously send data to Node B without Node B being able to determine where the data originated from.  When a TOR client is started, a list of address (IPs) and other information concerning the entire TOR network is received.  Based on this data, the TOR client decides upon a random path in the network. Let the path be:  A->D1->D2->D3->B.  Therefore a packet of information will travel between these nodes in order to reach the destination B.  Node A first encrypts the data with the key of D3, followed by an encryption using D2 and lastly an encryption using the key of the next hop, D1.  Encrypting data with a key of Z means that only computer Z can decrypt this data.  The packet is then sent from Node A to Node D1.  Node D1 decrypts the ‘first layer’ of encryption and then forwards the packet to the next hop which is contained in the decrypted data.  This process continues from D1 to D2 and from D2 to D3.  When D3 receives the packet, the final encryption is removed and the original unencrypted packet is left.  Node D3 then sends the data to Node B, the destination.  (The original packet must be modified such that the Origin is now of Node D3).  Node B now believes that D3 (end node) is the source of the data.

Due to the encryption, no TOR node (D1,D2,D3) knows the complete path of the packet.  Each node only knows the immediate hop before and the immediate  hop after it. When Node B tries to send a packet back to Node A, he will first send it to Node D3 who will forward the data to D2.  The encryption happens in reverse this time.

Implications:

TOR is becoming more popular with groups requiring anonymous posting of data and information.  Journalists can use TOR to safely communicate with informants and even government employees can use TOR while abroad to hide affiliations with a particular country.  While TOR can be used for good things it can also be used for spam and other harmful activities.

In a study at Kansas University conducted by Michael Vitevich, an attempt is made to map out the words of our vocabulary as they are represented in the human brain. What Vitevich hopes to learn from charting out our vocabularies in the brain is how victims of brain trauma regain their language skills, but this information can also be used to study many things including language development and processing. As it turns out, the idea of the “six degrees of separation” that was discussed in class shows up in these models. Most words are unusually close to one another, implying that the network must be extremely complex with many edges. The mappings themselves are done using the relation of words to one another based on their sound and meaning, as seen in the diagram below.

As mentioned in the article and seen from the diagram, there are several words that have many more edges than others. These words appear like the gatekeepers we have mentioned in class in the sense that removing, or “forgetting” one, will make it hard or impossible to connect or remember some words. For example, removing “peach” in the diagram will split the vocabulary of this person into two parts. Another related idea here is that of structural holes, which can be seen at the gatekeepers as well as other nodes that connect two large subsets of the graph. The structural holes can tell us a lot about the way humans think about things, and why certain words are clustered together. If researchers can learn where the key words, or gatekeepers, are in the word network, this can have major implications on helping children and brain trauma patients build up their vocabulary faster because these kind of words are essential for remembering many others. One can think of learning the word cat at a young age, and then immediately words like rat, fat, mat, etc. come along. Clearly, the network of words that our brains construct is a very complex system that relies on key words throughout, although finding the exact structure of the network still seems very fuzzy.

http://www.news.ku.edu/2008/february/5/language.shtml

http://www.slideshare.net/faberNovel/social-network-websites-best-practices-from-leading-services?src=embed

A recent study by FaberNovel (the results of which are published in slideshow form), a management consulting firm, highlights aspects of prominent social networking websites and point out some broad principles for understanding them and concludes with case studies on online matchmaking websites and business networking websites. While we haven’t discussed all of these principles, I believe they offer valuable insight into how social networking sites function in real life.

One of the interesting things that the study discusses is the fact that many online social networks are “scale-free networks”, which means that they are 1) organized around some central nodes, and 2) they grow through preferential attachment, which means that the more connections a node has, the more likely it is to receive even more connections. In broader terms, a node with many edges coming out of it, connecting it to other nodes (i.e. high degree) can be compared to someone who is well-known in a community. Then, whenever new members are introduced into the community, they are more likely to “hear” or “know of” and thus become acquainted with the well-known people than the not-so-well-known (nodes with low degrees). This is similar to the concept of Strong Triadic Closure in that both show the expansion of a network. They are different in that while Triad Closure ensures that the entire network (of the 3 friends, for example) grows evenly (i.e. two edges coming out of each node), preferential attachment allows the more prominent members in a community to grow even more. In a sense, triadic closure allows others to grow while preferential attachment allow you to grow.

http://tobolds.blogspot.com/2007/04/general-economic-equilibrium-in-world.html

Although the market of each server in World of Warcraft is very large, the market is simplified by allowing direct exchange between players in an auction house. The auction house is a place where players can put items they don’t want up for sale, with a starting bid and buyout price, for other players to search through and purchase. This article discusses the effect of gold farmers on the prices of items. Gold farmers are typically people who’s job is to hang around in a certain spot, collect one specific thing, sell that item on the auction house and sell the Warcraft gold they make to players for real life money. Players in general tend to use these gold farmers as scapegoats for any drastic increase in the price of a certain item, but this article seeks to clarify that this is not only untrue, but that the effect of gold farmers is actually lower prices.

Previously in class we have discussed market exchange theory. Although the auction house is technically an auction, the buyout price is almost always the selling price of an item. This allows us to model each item being sold as a seller with the seller’s valuation being the buyout price. There are no traders, and the buyers are the players who wish to purchase the item. Because of their hording nature, gold farmers will typically have more of a certain item than normal players. This will lower the gold seller’s valuation of each item relative to the average player. When this gold seller goes to put their items on the auction house, there is effectively several additional sellers added to the market with a lower item valuation. As we can see by working out this simulation on paper, this will result in a lower selling price assuming the number of buyers is less than the number of sellers. The buyers end up better off, and the sellers end up worse off. What we can see from this is that if there are gold sellers flooding the market with cheaper versions of the items you are trying to sell, the best way for you to make money is to find a market with less sellers, or at least where the sellers do not outnumber the buyers.