Techniques used to extract sentiment from social media data are complex, at times counterintuitive, and often laden with assumptions. Before providing readers with a how-to guide to implement these models, we think it is critical to explain the techniques in depth so users can deploy them appropriately. This chapter explains the theoretical grounds for the techniques developed in the next chapter and serves as a bridge between the discussion of the pitfalls of social media mining and the execution of that mining.
We find it useful to situate social media mining within the context of traditional social science research. While defining social science is difficult, Jean Anyon's perspective is a nice starting point. She suggests that socially explicit theory, and thus social science, should be empirically constructed, theoretically defensible, and socially critical. More generally, social science's main aims are to generate theories that explain individual-and group-level behaviors and then to examine the veracity of those theories with evidence. Generally, these theories are more valuable insofar as they allow a deeper understanding of human behavior, and especially so if they provide an understanding sufficient to allow for intervention. Our approach to social media mining strives to take this challenge to heart; thus, throughout this book, we use social media data to ask and answer questions of pressing social relevance.
Traditional social science not only focuses on important questions, but also seeks to uncover relationships that are interesting and unexpected. The world around us is full of complex social behavior; though identifying mundane facts is sometimes helpful in the name of basic research, it does little to help us understand social behavior. We take to heart the mandate to find interesting relationships as we mine social media data—a particularly complex and rich source.
At heart, however, social science is not a focus on the important or the interesting. It is science, which means that it is a set of methods and practices designed to generate and verify facts. The logic of science, regardless of whether it proceeds quantitatively or in a qualitative fashion, is fundamentally about knowledge discovery and accumulation. This logic helps mitigate several shortcomings in reasoning that frequently hinder our ability to make correct inferences. Some examples include illusory correlations (perceiving correlations that do not exist), selective observation (inadvertently cherry-picking data), illogical reasoning, and over or under generalizing (assuming that facts discovered in one domain apply to others as well). Generally, the scientific process helps avoid the discovery of false truths often arrived at through deduction, speculation, justification, and groupthink.