Lecture 1: Case Studies in Computational Social Science

This week’s lecture provided a high-level introduction to computational social science, an emerging field at the intersection of social science, statistics, and computer science that aims to use large-scale, individual-level data on who people are (demographics), what they do (behavioral & temporal data), and who they know (networks) to further our understanding of human behavior. In contrast to traditional approaches to social science (e.g., surveys, aggregate data analysis, and lab experiments), the large-scale nature of these data present unique computational and methodological challenges that will be discussed throughout the course.

We discussed a basic research loop, with three broad but important steps:

  1. Formulate the question.
  2. Find/collect data.
  3. Address methodological challenges.

While it may be tempting to think of this as a simple, linear process, in practice we often find ourselves iterating through the loop several times to address various complications that arise. For instance, we may need to refine the motivating question given methodological challenges or limitations of the available data, or develop novel methods to deal with computational issues.

We then discussed several different questions from a variety of domains:

  • Marketing: The long tail of consumption. What is the impact of inventory size on customer satisfication? How does interest in niche content vary across individuals?

  • Political science: The convention bounce. Do people actually switch which candidate they support? Are there overall population-level shifts in candidate support?

  • Demography: The digital divide. Does Internet access/usage affect health, education, and employment outcomes? How do Internet usage patterns vary across subpopulations?

  • Economics: Auction design. How do you optimally set auction parameters (e.g., reserve prices, “buy it now”, etc.)?

  • Communication theory: Information diffusion. How do ideas and products spread through society? What is the empirical structure of diffusion cascades?

The technical details of investigating these questions will be the subject of subsequent lectures, but curious readers can find more information in the following papers.