Graphs provide ways to study networks because these graphs show the nodes in a network, the edges between nodes and the tie strengths these edges have and if these edges have a certain direction. Graphs provide a picture to something we cannot readily see, but the reliability of a graph depends on the date gathered about a network. The more date used, the better the graph, or to put in different terms, big data produces big graphs. Big data is exactly as it sound, large amounts of data gathered and used to draw conclusions. The emphasis on big means that the data set is so large that it can only be analyzed by computers. Graphs are one of the perfect tools used to get a snap shot of large networks, networks so large we need to use computers to build these graphs.

Samuel Arbesman in his article “Stop Hyping Big Data and Start Paying Attention to ‘Long Data'” argues that the current focus of big data is too small. He states that the picture that big data shows us right now is too limited to truly draw important and far reaching conclusions. Arbesman wants big data to incorporate a larger time line, not just the past one hundred years, but the past century. This argument brings up an important limitation of graphs when studying networks: graphs are snap-shots of a network. When you look at a graph, not matter how much data you have put into it when studying a network, it only shows you as much as the most recent and oldest pieces of data you have at that moment in time.

Take for example, a graph of a social network, you can see friend A is friends with B and C, and these are very close friends with A, but friend B and C are not friends. In other words, node A has a strong tie to node B and node C, but node B and C have no edge; therefore node A violates the Strong Triadic Closure assumption in this graph. Now, say we look at this same social network in one month and find that friend C and B have met and now are friends, the strong triadic closure assumption is now satisfied. Applying this thought process to other networks made up of many more nodes makes graphs seem far too static to truly understand a network.

I think to truly appreciate the uses a graph has in studying networks, you need to look at a series of graphs about a network to draw conclusions, but, as Samuel Arbesman points out, we need to incorporate not the most recent data, but as much data as we can pull to form the best conclusions these graphs have in them. One graph is only worth the amount of nodes and edges in it, but multiple graphs are worth exponentially more in studying a network.