Social Network Theory

The Wayback Machine - https://web.archive.org/web/20041116125614/http://www.ccs.neu.edu:80/home/perrolle/archive/Ethier-SocialNetworks.html

Current Research in Social Network Theory

Jason Ethier

1�� Introduction

�� In this paper I will discuss some of the most state of the art research being done in the field of social network theory. I will briefly cover some of the technology and theory behind the research, but will mainly focus on the sociological implications. There are vast arrays of topics being studied in social network theory and this paper covers a range of the most important and interesting research. The study of social networks is important since it helps us to better understand how and why we interact with each other, as well as how technology can alter this interaction. The field of social network theory has grown considerably during the past few years as advanced computing technology has opened the door for new research. Before delving into the current research, I will present a brief introduction to the foundations of social network theory.

�� Social network theory is a branch of social science that applies to a wide range of human organizations, from small groups of people to entire nations. The term network refers to a set of objects, or nodes, and a mapping or description of the relationship between the objects. In the case of social networks, the objects refer to people or groups of people. For example, a network might consist of a person and a mapping from that person to each of his or her friends and relatives. These mapping can be directional or bi-directional. An example of a directional mapping would be if person A liked person B, but person B did not like person A. This is a directional mapping from person A to person B. An example of a bi-directional mapping would be if person A and person B both liked each other.

�� One of the reasons social network theory is studied is that by understanding the mappings connecting one individual to others, one can evaluate the social capital of that individual. “Social capital refers to the network position of the object or node and consists of the ability to draw on the resources contained by members of the network” [1]. Basically the more mappings a person has in the social network and the more mappings these people have, the more knowledge, influence, and power the original person will control. Social capital can have a substantial influence on a person’s life; affecting such aspects as job searches and potential for promotions. Social networks can also help sociologists identify primary groups and cliques. I will now discuss some of the current research in the field of social network theory.

2�� Social Network Theory and the Formation of Public Opinion

�� One of the questions researchers are working on is how social network theories can describe the formation of public opinions. Most researchers in this area are concentrating on the political power of social networks. The question of how networks influence political agency and behavior is of tremendous theoretical and practical interest. These researchers believe that collective action, voting choices, and other methods of political participation are controlled by social networks. They try to simulate collective processes of public opinion formation in order to better understand exactly how social networks influence politics. Researchers have developed models of how opinion changes occur in a network. “Actors increase their interest to participate in public processes if connected with others with higher interest levels who contribute and they decrease their interest if connected to others with a lower level who defect” [2]. In this way collective action occurs only if there is a positive correlation between interest and power. Thus a population having differing levels of interest is found to have positive effects on increasing the population’s potential for participation. Whereas populations in which all the participants share the same level of interest tends to stifle political participation.

�� Other researchers have developed a model of collective behavior in analogy to physical systems. In this model each actor possesses a strength factor of opinion and the probability of choosing an opinion is proportional to the number of actors who hold that opinion. Thus the likelihood of a group coming to a certain consensus depends on the group distribution of opinion. These researchers study what conditions are necessary for a group to change opinions and how this is dependent on the size of the group. They also study a form of social automata in which actors interact only with those in their vicinity according to some well-defined rules. The goal is to study any group patterns that emerge from this interaction according to fixed rules. What these researchers are most interested in are abrupt state transitions from group consensus to near consensus to nonconsensus within well-ordered pockets of opinion [2].

�� Citizen involvement in political institutions, and individual decision-making about voting and participation, is considered to depend on social psychological perceptions and beliefs, social forces impinging upon the citizen and social interactions among citizens. This suggests that there should be a relationship between social connectedness and political participation. A model of this behavior is one in which individuals are seen as parts of loosely knit, flexible networks in which information transmission occurs through political discussions. People form their opinions on the basis of the perceived quality of the information from individual discussions. This leads some researchers to believe that the formation of public opinion is like collecting the conclusions of thousands of individuals serving on different juries. In this view there are many small groups with a formed opinion, and there is much variance between the opinions of different groups. These differing small group opinions combine to form the overall group opinion. Thus in order to win an election, a candidate’s supporters must convince those with the most social capital within each small group in order to have the possibility of winning the support of the majority of small groups.

3�� Construction of Reputation for Network Members

�� Social networks can be especially important in the construction of a person’s reputation. This is especially apparent in online marketplaces such as eBay. eBay is an example of a large multi-user system where interpersonal communication between members is scarce. In systems such as this, it can be very difficult for members to build a reputation without the aide of specific tools for this purpose. Reputation can be defined as the common or general estimate of a person with respect to character or other qualities. This estimate is necessarily formed and updated over time with the help of different sources of information. Sociologists have been studying how social networks can be used to update and analyze trust and reputation. These studies show that it is possible to say a lot about the behavior of individuals using the information obtained from the analysis of their social networks.

�� Researchers have created a model to describe how reputation is determined in an e-commerce community. Their model considers three types of relationships between community members. These relationships are competition, cooperation, and trade. Competition is the type of relation found between two or more members that pursue the same goals and need the same scarce resources. Competition generally has a negative impact on the reputations of those involved. Cooperation implies significant exchange of sincere information between the members and some form of predisposition to help each other. Cooperation tends to improve the reputations of members who participate. Trade reflects the existence of commercial transactions between two agents and is compatible with either cooperative or competitive relations. Trade generally helps a member’s reputation but can also hurt it. This model also uses three types of social reputation depending on the information source. These are witness, neighborhood, and system reputations. Witness reputation is based on the information about a member coming from other members who share a relationship with that member. Neighborhood reputation uses the social environment of the member, that is, the neighbors of the member and their relations with it. System reputation is a default reputation value based on the relations the member is currently engaged in and has belonged to in the past. For example, those members who have consistent competition relationships will receive a negative system reputation value [3].

�� The use of the social network analysis techniques as part of a reputation system opens a new field for experimentation. Once you introduce the social dimension in reputation evaluation and the members start to take into account social relations, it becomes more and more important to consider not only which is the reputation of the other members, but also what can a member do to get and maintain a good reputation. Efficient methods of evaluating reputations can lead to more hospitable relations among members of the community. Users may be less inclined to enter competitive relations when they know the competition may hurt their reputation. The information researchers have gathered can be used to improve current methods of reputation evaluation on e-commerce websites.

4�� Power and Stability Within Groups

�� Recently there has been considerable research on the topics of power within social networks and the stability of networks. Stability is determined by the likelihood of members leaving one group for another due to dissatisfaction with the members of the original group. The first major question researchers are studying is: what characteristics are associated with stable networks? What researchers have found is that a balance of power within a social network is necessary for stability within the network. These researchers conclude that only strong power networks are unstable. A strong power network is characterized by some members owning complete power at the expense of other members. This contrast in power levels causes friction between members of the network and will eventually lead to instability. This social friction is avoided in networks where each member shares a relatively equal amount of power. People are more likely to stay in a group where they share equal power with their peers [4].

�� Another important question asked by these researchers is: if members of a network migrate when they are dissatisfied with their power, is the result inevitably a network of equal power? The answer researchers have found to this question is no. Groups where members have been divided into a hierarchy of power tend to continue this process of division of power even if the group becomes unstable. Their research has shown that, surprisingly, as the group becomes small enough that those members with the highest power can no longer exert their power on weaker members, a new hierarchy of power will emerge. This research has implications for the study of social networks in politics, networks in the workplace, as well as networks and discrimination. Researchers are currently working on exactly how power is first introduced into the social network in order to obtain a better grasp of the entire evolutionary process of stability and power within social networks.

5�� Groups as Dynamical Systems

�� Other researchers have attempted to describe the behavior of small groups as complex or dynamical systems. This presents a different method of inquiry for the study of groups. An overview of this method is as follows. In small groups local action consists of recursive, nonlinear interaction among many different people or elements. Local group process creates, activates, replicates, and adjusts dynamic links in a coordination network. This can be thought of as an interaction among many local variables. From local action, patterns on the global level emerge. These are behavioral and cognitive patterns such as group norms, cohesion, division of labor, a role system and influence structure, and temporal patterns such as cycles of conflict and consensus. These global patterns can be thought of as global variables that emerge from local interaction and then structure subsequent local action.

�� Local action for any given group shows regularities or patterns, which can be modeled as a set of rules that the system follows. The interaction among local level elements may be highly complicated; however the rules governing the action and interaction of group elements are often relatively simple. The researchers believe that the rules guiding local action and which global patterns emerge from the operation of these rules, depends on initial conditions. The entire pattern of global dynamics that emerges from this local action may shift when one of these initial conditions is slightly altered.

�� Given the range of potential interactions among local variables, it is not possible to predict the individual and joint values of these variables accurately, even if their values are known with high accuracy at a particular point in time. Other complex systems, such as the weather, whose behavior depends largely on interactions among local elements, are predictable only in the short term. These predictions are for global variables such as overall weather patterns, not local variables such as the exact path of a tornado. However, patterns of key global variables do show substantial regularities over time. One similarity of almost all dynamical systems is that global variables settle over time into relatively small regions of possibilities for that variable. If these regions can be identified in the study of social networks, it would greatly enhance the predictive capabilities of social network theory. These researchers are attempting to track the characteristics of social networks through different states, as reflected in the pattern of global variables over time. If they are successful it would not only provide them with a better understanding of how certain factors influence social networks, but it would also allow them to better predict the behavior of social groups of all sizes [5].

�� Knowledge of human networks can also be applied to the design of computer networks. Computer networks are put in place to support human networks; person-to-person exchanges of information, knowledge, ideas, opinions, insights, and advice. Researchers are working on ways to apply the algorithms of social network analysis to designing computer networks. Social network analysts look at complex human systems as an interconnected system of nodes (people and groups) and ties (relationships and flows) much like an internetwork of routers and links. Human networks are often unplanned, emergent systems. Network ties often end up being unevenly distributed, with some areas of the network sparsely connected. These are called small world networks. Computer networks often end up with similar patterns of connections, dense interconnectivity within subnetworks, and sparser connections uniting subnetworks into a larger internetwork.

�� Human networks, like computer networks, can be subject to the limitation of a single point of failure. In human networks people at times can play the role of broker or connection between two different smaller groups. If this person changes their location in the human network, the two groups will no longer be able to communicate and the human network may start to break down. One of the most important ways to obtain social capital in a human network is to have the shortest path to as many others as possible. Maximizing closeness between all routers improves updating and minimizes hop counts. The shorter the path, the fewer hops/steps it takes to go from one node to another. In human networks, short paths imply quicker communication with less distortion. In computer networks many short paths connecting all nodes will be more efficient in passing data and reconfiguring after a topology change [6].

�� In a recent network design book, Advanced IP Network Design, the authors define a well-designed topology as the basis of a well-behaved and stable network. They propose the idea that, “…three competing goals must be balanced for good network design: reducing hop count, reducing available paths, and increasing the number of failures the network can withstand” [7]. Social network algorithms can assist in meeting all three of these goals. Reducing the hop count infers minimizing the average path length throughout the network. This can be done by maximizing the closeness of all nodes to each other. Reducing the available paths leads to minimizing the number of shortest paths between members in the network. Increasing the number of failures a network can withstand focuses on minimizing the centralization of the entire network. Social network models can model our computer networks and suggest link changes to form an effective topology that has a short average hop count, not too many paths, and just enough redundancy.

7�� Social Networks and Personal Health

�� Recently there has been research into the study of the implications of social integration for personal health. This research has shown that participation in a diverse social network may have an influence on health. The researchers chose to study social network diversity (number of social roles) and susceptibility to the common cold in people experimentally exposed to a cold virus. What they have found is that the greater the social diversity of the person, the lesser his or her susceptibility to infectious illness will be. Despite these results, the researchers were not able to isolate the pathways through which social diversity was associated with susceptibility. The leading hypothesis is that as social diversity increases, the level of exposure to a certain illness also increases. Thus the immune system is better prepared to defend itself against any future exposure to the sickness. However, the researchers have so far not been able to thoroughly support this hypothesis experimentally. What this research does show is another strong benefit of having high social diversity or social capital [8].

�� The results found by these researchers are quite surprising, “The magnitude of the health risk of being relatively isolated (socially) is comparable to the risks associated with cigarette smoking, high blood pressure and obesity and is robust even after controlling for these and other traditional risk factors” [8]. It appears that cultural isolation can have a profound effect on physical well being. Their research has also shown that the development of mental illness is associated with the level of social contact a person has. Some researchers believe that this is due to the fact that people’s identities are tied to their social roles. By meeting role expectations, individuals are given the opportunity to enhance their self-esteem. They believe that these social roles provide a purpose to life. They imply that a sense of purpose is an integral component of psychological well being.

�� The social network measures used in studies of health outcomes are not as advanced as those involved in formal social network analysis. A major reason for this is that studies of health outcomes typically involve large samples and include multiple questionnaires or interviews. For these studies, intensive quantitative measurement is reserved for the rare cases in which the researcher determines that there exists sufficient need for it. Thus the social network results in this type of research do not always hold up to the same academic rigor as other research in the field of social networks does. This does not discredit the research described above. However, it does propose that further research is required before these conclusions can be adequately supported.

8�� Marketing Through Social Networks

�� Research in social networks has also proven to provide great benefits to the field of marketing. Social networks and their patterns of relationships are a fundamental fact of market behavior and can be used effectively as a basis for marketing strategies. A major challenge facing marketing strategists is how to increase the effectiveness of social network based marketing strategies. In order to reach this goal marketing researchers and scientists have collected social network related data and have analyzed it using social network analysis. The study of social networks is beginning to be widely used in marketing. One of the reasons why it has taken so long to have an impact is because of the scarcity and difficulty in obtaining the requisite data.

�� “Network marketing entails distribution of products and services through a network of independent businesspeople, who in turn either take care of the distribution themselves or recruit others to do so” [9]. This is one example of using social networks for the purpose of marketing. Current research is focusing on which types of people this form of marketing should focus on. Marketing strategists are not only looking for people with the most social capital, but also people who are associated with others who have access to a large amount of social capital. Marketing through social networks aims to take advantage of the social capital of each person who participates. An example of this is MCI’s “Friends and Family” campaign of the 1990’s. This plan offered discount calls to residential customers when dialing a telephone number from a list pre-specified by the customer. The customer in turn must furnish MCI with the names and other information about the people on the list. These people are then contacted by a sales representative from MCI in an attempt to induce them to enroll in the plan.

�� The main questions for researchers in this branch of social network theory are which types of social networks can be used as a basis for marketing strategy, how to identify and measure social networks, how to mobilize and manage social networks, and which marketing decisions can benefit the most from social network concepts and methods? Some researchers have applied the use of supercomputers to simulate the performance of marketing geared towards social networks. This technique has helped researchers to answer the above questions. What they have found is that consumer networks that are not under the control of a corporation work best for marketing purposes. An example of such a network is a word-of-mouth communication network in which people recommend a product to others within their social network. Corporations identify and measure social networks by collecting information from their customers. One method of doing this is by distributing discount cards in exchange for customer information. This field of social network theory is certain to be subject to increased research as more companies learn of the marketing potential of social networks.

9�� Social Networks and the Internet

�� Researchers at Stanford University have analyzed an online community at the school known as Club Nexus, in order to determine how the community reflects the real world community structure within the student body. Club Nexus allows users to chat, organize events, share opinions and photographs, buy and sell used goods, make announcements, and meet new people. The club has over 2,000 student members, comprising more than ten percent of the student population. Each member has a profile describing themselves and a list of buddies. One advantage of studying online communities is that they allow researchers to gather data with considerably less effort than other forms of communities. The researcher’s ability to learn more about the social network is simply a side effect of users transmitting information digitally.

�� One interesting tool provided by Club Nexus is the ability to send messages and invitations to a certain degree of connections in the social network. For example, members can send a message to the people listed in their buddy list, or the buddies of each person listed in their buddy list, or the buddies of those buddies, etc. This is one way of using social networks to communicate with people whom you may not know directly. Researchers found that the average distance between any two members (measured in the number of hops along the Nexus network) is only four on average [10]. This result is very interesting considering that Club Nexus represents a diverse group of users, both undergraduates and graduates, at various stages in their studies, representing many different departments.

�� The researchers analyzed correlations between the profiles users provided and connections between these users in the social network. This analysis was able to detect some expected trends, such as people sharing narrow or unusual interests were likely to become friends. It also uncovered some non-obvious relationships, such as people who described themselves as being ‘responsible’ being perceived as slightly less ‘cool’ by other members of the network. What makes online communities such as Club Nexus unique is that one is able to observe these patterns on a large scale with many different variables. The richness of this information can be used to model dynamics such as the spread of ideas on a network or the way that people can find each other through their contacts. Researchers are now studying how this online community evolves over time and how social dynamics, such as the adoption of new features introduced to the web site, affect the community.

�� In the past few years many social networking tools such as Club Nexus have appeared online. Some of these tools created by companies such as Linkin, Ryze, Friendster, Spoke and VisiblePath, are attempting to profit from the networking capabilities they provide. When the usefulness and potential profitability of these applications of social networking are evaluated, we must consider the possibility that the growing bubble around social networking applications may be about to burst. It’s not that social networking applications do not have the potential for future benefits. Their main pitfall is that venture capitalists are pouring money into companies creating these applications even though many do not have concrete business models. These companies may be falling into the same trap as many of the dot coms have.

A student at Caltech has created a website (BuddyZoo.com) that applies social network analysis to AOL Instant Messenger (AIM) buddy lists. Users submit their AIM buddy lists to the site and BuddyZoo runs various forms of analysis on the data. BuddyZoo allows users to find out which buddies they have in common with their friends, measure how popular they are, detect cliques they are a part of, see a visualization of their buddy list, and see the degree of separation between different screen names. The degree of separation feature allows users to determine the shortest path from their buddy list to another person’s screen name (i.e. how many different people’s buddy lists they would have to go through before reaching the specified screen name). Users can also view their prestige level, which is computed in a similar way to the method Google uses to compute page rank for web pages. This prestige level is similar to the social capital concept discussed earlier. BuddyZoo currently has a database of 7,936,710 unique screen names that it uses for analysis [11]. This is a good example of an interesting approach to the study of how technology supports the growth of social networks.

�� All of the companies listed above have used the Internet as their means of generating virtual social networks. The Internet is certainly an amplifier for this sort of social interaction. It is used as a solution to the social networking problem of how to close the gap of separation between people around the world. Companies are attempting to find the shortest path to a person, whether they are trying to sell a product, find a date or locate an old friend. There are three major difficulties involved with the current social networking solutions available online to solve problems of this sort. “Perhaps the biggest barrier to social networking solutions’ usefulness is critical mass: getting enough people to join the network so that people can find each other” [12]. Unless there is a relatively large body of participants socializing by using the application it just won’t be successful. The hardest part of recruiting members is getting the initial group to join. Once a small but committed membership group has been established the size of the group will begin to grow exponentially as those who use the network bring their outside network of friends into the group.

�� There are two other major problems social networking solutions must overcome. The first being that it is just too much work to upload your contacts into all the various social networking applications. Actively participating in more than one or two of these applications consumes far more free time than the average member has available. The other major barrier to the productive application of social networking systems is that they are being developed as standalone systems instead of being incorporated into the information technologies that businesses are already using to manage business relationships or relationship-related information. Creating social networking tools that extend solutions already in use instead of making people use third-party applications is essential. This last problem applies only to enterprise market social networking solutions. This is where experts believe the future applications for the social networking market will be [12].

10�� Intelligence Applications of Social Network Analysis

�� Research in social network analysis is being performed by government agencies for use in defense programs. The Total Information Awareness program sponsored by the Defense Department is currently working on a project known as Scalable Social Network Analysis (SSNA). “SSNA aims to model networks of connections like social interactions, financial transactions, telephone calls, and organizational memberships” [13]. They are attempting to model the social networks that terrorists belong to. The purpose of the SSNA algorithms program is to extend techniques of social network analysis to assist with distinguishing potential terrorist cells from legitimate groups of people, based on their patterns of interactions, and to identify when a terrorist group plans to execute an attack. This is an extremely ambitious project considering the scale of the social networks that these researchers are attempting to model. In order to be successful SSNA will require information on the social interactions of the majority of people around the globe. Since the Defense Department cannot easily distinguish between peaceful citizens and terrorists, it will be necessary for them to gather data on innocent civilians as well as on potential terrorists.

�� The SSNA program is developing techniques based on social network analysis for modeling the key characteristics of terrorist groups and discriminating these groups from other types of societal groups. Social network analysis (SNA) techniques have proven effective in distinguishing key roles played by individuals in organizations and different types of organizations from each other. For example, most people interact in several different communities. Within each community people who interact with a given individual are also likely to interact with each other. According to the Defense Department very preliminary analytical results based on an analysis of the Al Qaeda network of September 11^th hijackers showed how several social network analysis metrics changed significantly in the period immediately prior to the attacks. This change could have indicated that an attack was imminent.

�� Current SNA algorithms are effective at analyzing small numbers of people whose relationship types are unspecified. SSNA would extend these techniques to allow for the analysis of much larger numbers of people who have multiple interaction types (i.e. communication and financial). The program will develop algorithms and data structures for analyzing and visualizing the social networks linkages, implement these algorithms and data structures into software modules that provide SNA functionality, and demonstrate and evaluate these models in appropriate intelligence community systems and environments. SSNA begins development in fiscal year 2004 and is expected to conclude in fiscal year 2007. It is generally regarded as one of the most ambitious projects in social network analysis ever attempted [14].

�� We have seen how the study of social networks is currently being applied to many different fields. The volume of research has dramatically increased as corporations realize the importance of social networks as both a marketing tool and as a means of communication between customers and employees. The implications of social network theory extend beyond the applications of business to explain the hierarchy of social and political power that exists in our society. A person’s social network can affect them in a variety of ways, from their reputation to their health. Social networks are dynamic and evolve to fit new technologies that are introduced to society. The Internet has allowed social network interaction to expand in ways that were previously not possible.

�� As our technological potential advances our society will have to determine whether social networks should be subject to the same privacy protections that apply to individuals. As governments and corporations begin to understand the volume of information that can be obtained by monitoring network interaction, we will begin to see more and more applications created to map our social groups. For the purpose of indirect marketing and intelligence, these applications will be used to learn whom we interact with on a regular basis and exactly how the interaction is performed. Governments and organizations will begin to see how the currently existing network of social groups can be used as both a tool for generating profit and creating and sustaining positions of power. The study and application of social network theory is certain to continue to be a field of considerable interest in the future.

[1]�� Kadushin, Charles. “A Short Introduction to Social Networks: A Non-Technical Elementary Primer” Accessed: October 29, 2003. Online. Internet. <http://construct.haifa.ac.il/~cerpe/papers/kadushin.html>.

[2]�� Moses Boudourides(2002). “A Review of Network Theories On The Formation of Public Opinion”. Contributed paper at the EURICOM Colloquium: Electronic Networks and Democratic Engagement, Nijmegen, the Netherlands, October 9-11, 2002.

[3]�� Jordi Sabater, Carles Sierra(2002). “Reputation and Social Network Analysis in Multi-Agent Systems”. International Conference on Autonomous Agents. July 2002, pp. 475-482.

[4]�� Phillip Bonachich(2001). “The Evolution of Exchange Networks: A Simulation Study”. Journal of Social Structure. Volume 2, number 5, November 8, 2001.

[5]�� Holly Arrow, Joseph McGrath, and Jennifer Berdahl(2000). “Small Group As Complex Systems”. Sage Publications, Inc., London.

[6]�� Valdis Krebs(2000). “The Social Life Of Routers”. Internet Protocol Journal. Volume 3, Issue 4, December 2000.

[7]�� Retana, A., Slice, D., White, R., Advanced IP Network Design, ISBN 1578700973, Cisco Press, 1999.

[8]�� Sheldon Cohen, Ian Brissette, David Skoner, and William Doyle(2000). “Social Integration and Health: The Case of the Common Cold”. Journal of Social Structure. Volume 1, number 3, September 5, 2000.

[9]�� Phipps Arabie, Yoram Wind(1994). “Marketing and Social Networks”. Sage Publications, Inc., London.

[10]�� Lada Adamic, Orkut Buyukkokten, Eytan Adar(2003). “A social network caught in the Web”.

[11]�� Accessed:� December 2, 2003. Online. Internet. <http://www.BuddyZoo.com>.

[12]�� “The Promise and Pitfalls of Social Networking”, Accessed: December 1, 2003. Online. Internet. <http://www.darwinmag.com/read/110103/pitfalls.html>

[13]�� “EFF Review Of May 20 Report On Total Information Awareness”, Accessed: December 1, 2003. Online. Internet. <http://www.eff.org/Privacy/TIA/20030523_tia_report_review.php>.

[14]�� “Report to Congress regarding the Terrorism Information Awareness Program”, In response to Consolidated Appropriations Resolution. May 20, 2003. Pub. L. No. 108-7, Division M, 111(b).

Oct	NOV	Apr
	16
2003	2004	2006