Online interaction is now a regular part of daily life for a demographically diverse population of billions of people worldwide. Facebook and Twitter are two of the most popular places where these interactions take place. A key difference is that most content on Twitter is publicly accessible via the Twitter API or through data resellers such as GNIP and Datasift, whereas most Facebook content is private. Thus, Twitter has emerged as the single most powerful “socioscope” available to social scientists for collecting fine-grained time-stamped records of human behavior and social interaction at the level of individual events. These data are also global in scale, allowing researchers to address fundamental questions about social identity, status, conflict, cooperation, collective action, and diffusion. This unprecedented opportunity comes with a number of methodological challenges, including generalizing observations to the offline world, protecting individual privacy, and solving the logistical challenges posed by “big data.” This introductory chapter reviews current advances in online social research and critically assesses the theoretical and methodological opportunities and limitations.

Opportunities and Challenges for Online Social Research

Scientific disciplines make revolutionary advances not only through new discoveries, theories, and paradigms, but also because of the invention of new tools and methodologies (Kuhn, 1962). The electron microscope, space telescope, particle accelerator, and magnetic resonance imaging (MRI) have allowed scientists to observe the world at greater scale or at finer resolution, revealing previously obscured details and unexpected patterns and experiencing the “eureka moments” of scientific breakthroughs.

Twitter has been widely used for human behavior research with its applications in public opinion mining (cf. Chapter 3), studying well-being (cf. Chapter 4), disease monitoring (cf. Chapter 5), and disaster mapping (cf. Chapter 6). In this chapter, we review existing work on using Twitter data to measure socio- economic indicators, including unemployment rate, consumer confidence, social mood, investor sentiment, and financial markets. Moreover, to complement research with Twitter data, we use several examples to illustrate the use of other large-scale data sources (e.g., web search queries, mobile phone calls) for socioeconomic measurement and prediction. At the end, we discuss challenges with existing research and identify several directions for future work.

Introduction

There has been considerable success in leveraging large-scale social media data at the intersection of social sciences and computational sciences with myriad applications in socioeconomic measurement and prediction.

An early study (Antweiler & Frank, 2004) finds that the message volume of stock message boards on Yahoo! Finance and Raging Bull can predict market volatility, and disagreement among posted messages is related to high trading volume. Public mood indicators extracted from social networks such as Facebook (Karabulut, 2011), LiveJournal (Gilbert & Karahalios, 2010), and Twitter (Bollen, Mao, & Zeng, 2011) can predict stock market fluctuations. Zhang, Fuehres, and Gloor (2010) study the correlation between emotional tweets and financial market indicators. They find that the percentage of emotional tweets is negatively correlated with Dow Jones, NASDAQ, and Standard and Poor's (S&P) 500 values, but positively correlated with Volatility Index (VIX). Bollen, Mao, and Zeng (2011) develop a multidimensional mood analysis model that can track Twitter mood in six dimensions (i.e., calm, alert, sure, vital, kind, and happy) and find that Twitter calmness has significant predictive power on daily Dow Jones Industrial Average (DJIA) price changes.

Twitter, crowdsourcing, and other medical technology inventions producing real-time geolocated streams of personalized data have changed the way we think about health (Kostkova 2015). However, Twitter's strength is its two-way communication nature – both as a health information source but also as a central hub for the creation and dissemination of media health coverage. Health authorities, insurance companies, marketing agencies, and individuals can leverage the availability of large datasets from Twitter to improve early warning services and preparedness, aid disease prevalence mapping, and provide personal targeted health advice, as well as influence public sentiment about major health interventions. However, despite the growing potential, there are still many challenges to address to develop robust and reliable systems integrating Twitter streams to real-world provision of healthcare.

Introduction

Health and well-being are top priorities for citizens, societies, and governments from individual to national and international levels. In particular, public health threats have become a focus due to globalization, the emergence of new diseases, and the reappearance of older infectious diseases. The SARS outbreak in 2003 illustrated how quickly a new virus could spread globally.

The growth of the Internet and social networks has enabled unprecedented support for individual health monitoring and personalized care, as well as large-scale public health measures (Smith & Christakis, 2008; Valente, 2010; Paul & Dredze, 2011). In particular, the role of Twitter in the health domain now encompasses a two-way channel for managing health and seeking health advice at the personal level on one hand, and for aiding early warning and out-break response systems with the aim of mobilizing knowledge and scientific expertise to protect populations on the other hand.

The spectrum of Twitter health research outreach is impressive. Paul and Drezde (2011) investigated how Twitter became an important medium for understanding concerns about public health, while Christakis and Fawler (2007, 2008) looked into the dynamics of health behavior on a large social network investigating smoking and obesity.