Saturday, September 24, 2016

Thoughts on Watts' Obvious: A perspective on modern, data-rich social science, intuitive to a natural scientist

In Everything Is Obvious, Duncan Watts provides an overview of the emerging new field of computational social science and how the ability to collect large-scale information on behavior is potentially changing the way we approach social science.

He starts off by describing our strong intuition for social situations -- perhaps in contrast to what we have for the physical ones. This intuition (basically, common sense) sometimes is valuable but sometimes gets in the way. We can easily fool ourselves in social situations. For example, in the famous Lake Wobegon effect, everyone ranks himself or herself above average. Also, there is the way context often distorts how people perceive the relative merits of something; this is known as anchoring. For instance, an anchor price in negotiation often affects what people eventually pay.

Watts then talks about how with web searches, we are able to collect vast quantities of information on social behavior and do many controlled experiments on people. An example of this is an A/B type of experiment where one group is presented with one type of ad and another group is presented with a different type; one can really see if the ad makes a difference.

These types of experiments have the potential to be a new information gathering tool for social science, in a sense analogous Copernicus' telescope for the physical sciences -- providing huge amounts of hard data that we can theories against. In general, Watts is continuously comparing the situation in the social sciences with those in the physical and biological sciences. He talks about how different disciplines operate at different scales. Often moving up from one scale to another brings about a new set of phenomena that are difficult to predict from the previous scale. Hence, chemistry deals with a different scale than physics, biology moves up a scale again (atoms to molecules, to cells and then organs). Then when we get into the social sciences we look at individuals and then whole collections of individuals. In each case it is more and more difficult to rationalize the next scale. As we jump scales we see emergent properties coming about from the system and this is perhaps one of the more interesting things we observe in complex systems.

Watts talks a bit about how in the simpler systems one has very complex models and lots of data to test against and one can make an exact prediction whereas in more complicated situations one cannot make a prediction per se about a particular event but just a statistical generalization that something will happen a given fraction of the time.

Watts gives a nice overview of some of the research in social science for a nonsocial scientist, providing an overview of recent work in network science and famous studies such as Milgram's six degrees of separation. It turns out, as we all know, everyone is fairly well connected.

Overall, this book is a great read for someone with a bit of a physical science perspective trying to grasp social sciences. It is easily accessible and presents a lot of interesting ideas about collective social phenomena in terminology that other scientists can readily grasp.

TAGS: obvious0mg
private mail tags:

Everything Is Obvious: Once You Know the Answer
by Duncan J. Watts (Author)