Rianna Herzlinger
The term “End of Theory” first emerged in an explosive WIRED article, published by Chris Anderson in 2008. In this article, Anderson describes us as having entered the Petabyte Age: an era in which we have access to massive amounts of data (also called Big Data). As a result, the companies that leverage this data no longer know why phenomena occur— it is enough to know they occur. This extends to predicting human reactions to advertisements, creating translation systems, and matching ads to content. Humans have become more predictable, but arguably not more understandable. This attitude has extended to scientific innovation too. Where the traditional science involved a hypothesis, model, and testing, this process has become obsolete. As Anderson writes, “Petabytes allow us to say: ‘Correlation is enough.’” Thus, the end of theory refers to an era in which scientific hypothesizing and modeling is subordinated to the collection and analysis of massive quantities of data.
Is this a good thing? Anderson himself appears enthusiastic about our future without theory. He ends his article with the chilling statement: “There’s no reason to cling to our old ways. It’s time to ask: What can science learn from Google?” Part of his criticism of ‘the old ways’ is that the scientific models we relied on were really simplifications of phenomena, rarely correct but still somewhat useful. These models must be constantly refined when scientists encounter new facts about the world that violate their models. Anderson sees this as a cumbersome and unhelpful process— especially when the alternative is easier and often more accurate. Part of the explosive nature of this article is not Anderson’s identification of this trend, but his want to fully embrace a post-theory world.
My series will assess several examples of end of theory cases and discuss the hidden costs to reveling in Big Data at the price of theoretical knowledge of the world. Through these examples, I will track several themes. First, is the value of theoretical knowledge. Anderson makes a valuable point about models: they aren’t strictly true. They are necessary simplifications of the real world. But does that make them invaluable? I argue theoretical knowledge is still incredibly valuable. Second, is the interaction of machine learning (ML) models with theoretical knowledge. Often termed ‘black boxes’, these systems are seemingly inexplicable as to why they make the accurate predictions they do. Can we leverage ML to create new theories? I argue we cannot distill meaningful theory from the results of these models. Finally, I am interested in the consequences of a civilization that lacks understanding of their own world. To embrace the end of theory is to advocate for prediction sans understanding. What value does understanding bring?
Do simplified models actually help us understand the world?
Can we create theory from ML model results?
What are the consequences of a civilization without understanding of world phenomena?
This Case Study is part of a larger essay, Science, Data, and Knowledge: End of Theory and Understanding the World.
__________
Anderson, Chris. “The End of Theory: The Data Deluge Makes the Scientific Method Obsolete.” Wired, Conde Nast, 23 June 2008, www.wired.com/2008/06/pb-theory/.
Graham , Mark. “Big Data and the End of Theory?” The Guardian, Guardian News and Media, 9 Mar. 2012, www.theguardian.com/news/datablog/2012/mar/09/big-data-theory.
