 In the previous video, we looked at social networks on the micro-level, talking about individuals and local clusters. In this section, we'll be adopting a more macro-level perspective that focuses on the overall network structure within which individual actors are embedded. This top-down perspective we'll be following in this video seeks to understand and describe whole populations by the makeup of the overall network. Many among our considerations here will be the network's overall density of connections, its average path length and degree distribution. Overall network density is a primary determinant in the makeup of any network. With density, we're simply asking how many connections are there relative to the maximum possible number of connections. Network density can be understood in terms of interaction cost. The easier that it is for an agent to make a connection, the more connections we're likely to have. Network density is a very fundamental parameter in that it defines the difference between a social system that is a network as opposed to being a simple group of people. At a low level of connectivity, we're dealing mainly with individuals in isolation. Here it is the attributes and properties of those individuals in isolation that really matters. When we turn up the connectivity, this is no longer so much the case, it's now the nature of the network that you're a part of that really matters. This is captured by the famous saying, it's not what you know, but who you know. If we're dealing with a social system with a low level of connectivity, then it does matter what you as an individual know. But at a higher density of connections, it is more what your network knows that comes to matter. When we see this with the internet, we ourselves used to have most of the information and knowledge that we would use on a daily basis, but now much of the information that we use, we do not have ourselves, but it is instead in the network. Thus, as we turn up the degree of connectivity within the social system, it is no longer the attributes of the agents in isolation that is so important, but instead their capacity to interoperate provides something of value to the network and ensure their connectivity to that network. As another example, we might think about the difference between so-called introverted and extroverted people. Introverted people with a low level of social connectivity have to rely heavily on their own capabilities and they're often more self-resourceful than others, whereas extroverts who can rely heavily on their social network may not have such personal capabilities, but instead are particularly good at accessing the skills and resources that they themselves lack through their social network. Overall connectivity is then a primary determinant within any social system, and also one of the major determinants to the nature of power within that organization, as within a social system that has a very low density and loose coupling, not much power can be exerted. In high density social systems, there are more and stronger channels through which power can be exerted. Network density is also a key determinant to average path length. Here we're talking about how close two agents within the network are to each other on average. This closeness is obviously a very important factor in terms of network cohesion and interdependence. As we scale up the number of components to the social system, this creates longer path lengths between members. This can stretch and break traditional forms of social cohesion. A longer, average path length is like an outward force disintegrating the social system as it puts people at longer distances from each other with a lower sense of interdependence. We may also note the longer the path length, the easier it is for subgroups to form and disintegrate the overall social network. Agents act and adapt to their local environment. If we turn up the average path length between agents or groups, they will not identify with or adapt to those other members further away from them, and we may get the formation of incompatible local clusters. Trying to achieve global coordination within such a system would likely mean having to impose it in some top-down fashion. But now if we turn down the average path length, which might happen through better transportation or communications technologies, people now interact more often, making it easier for them to synchronize their states and easier for them to recognize their interdependence and common identity. Probably the second most important question we can ask about the overall structure to a social network relates to its degree distribution. Degree distribution is a measure of how evenly or unevenly the degrees of connectivity are distributed out among the agents. It is asking the question, do some people have lots of connections and others have very few, or does everyone have roughly the same degree? This is clearly going to tell us a lot about the nature of power within that system. A high degree distribution will mean inequality of some kind that will be detrimental to social integration, and it is in many ways this inequality and connectivity that is the means through which power can be exerted. This degree distribution tells us a lot about how centralized or distributed the overall network structure is. At a low degree distribution, all actors have relatively the same amount of connections, thus they would be what we consider peers, and we would get many peer-to-peer interactions giving us a distributed network. As an example of a distributed social system, we might think about the Israeli Kibbutz, which are collective communes in Israel that are traditionally based on agriculture. Within the Kibbutz, the principle of equality was taken very seriously up until recently. Members did not individually own tools or even clothing, they ate meals together in communal dining halls, and major decisions about the future of the community were made by consensus or by voting among all. Distributed social systems like this have limited centralized institutions, everyone is responsible for maintaining the system and power is thus distributed out. Although distributed social networks may exist, they are often the product of some random process or a small informal network, or they may be a network in its early formation where it has not yet developed any overall formal organization, or equally as in this example of the Kibbutz, the social network may be specifically designed to be egalitarian in nature. But more often what we see is that as a social network develops, and particularly when it becomes more formal, we get greater differentiation between degree distribution. Many real world social networks have a skewed node-degree distribution, in which most nodes have only few links, but by contrast there exist some nodes which are extremely well connected. This heavy tailed distribution is known as a power law or scale free network. Here we're getting the emergence of major hubs and high degrees of social inequality, and there may be two different reasons for this inequality. Firstly, some people are simply better at doing certain things than others. We all watch some people play football, sing or act simply because they're better at it than others. And what we mean by that is that they provide us with a better return on our investment of time, energy or money, and thus many of us choose to make connections to that particular node, while others do not receive our attention, thus giving us this unequal centralized model, and this process is meritocratic in nature. This explanation is largely intuitive to us, but it might not be suffice to explain how we can get such extreme differences in connectivity within these scale free social networks. Researchers have then also come up with another explanation behind the formation of these scale free networks, that of preferential attachment. As we've previously discussed, a preferential attachment process is any of a class of processes in which some quantity, typically some form of wealth or attention is distributed among a number of individuals or objects, according to how much they already have. So those who already have lots receive more, and those who already have little will receive less. One of the best examples of how preferential attachment works is seen in recent research done by Duncan Watson team, where they created two different websites selling music tracks, one where people could rate the songs that they downloaded and one where they could not. Over 14,000 participants then downloaded previously unknown songs on both sites. On the site where users could leave feedback for each track and others could see that feedback, it was found that there was a much greater disparity between the most downloaded song and the least compared to the other side where there was no feedback available. Thus increasing the strength of social influence increased the inequality in degree distribution. This power law distribution also applies to cities, the distribution of wealth and income and many other phenomena where we have social interaction, creating feedback loops that amplify the disparity to give us a much greater degree distribution than would naturally occur if simply generated by merit. And this creates major centralized hubs within the network, whether we're talking about urban networks, financial networks or some other social network. In this module, we've been talking about three of the major factors shaping the overall makeup to a social network. We started by talking about the density of connections as a primary factor, as it defines whether we're actually dealing with a network or just a group of independent people. And this level of overall connectivity will fundamentally change the whole system as when we turn it up, our focus has to shift from the properties of the parts to the flow of resources within the overall network. We then talked about average path length as a second key overall metric, one that will tell us a lot about the network's overall cohesion. Lastly, we looked at degree distribution as playing an important role in defining the degree of equality within the system. Degree distribution tells us a lot about how centralized or distributed the network is, which is of major significance in understanding the dynamics of power and how something will flow through the whole social system.