About This Site
Literature Search Procedure
Our literature search used two basic steps: We first generated a large pool of potentially relevant articles and then selected a smaller subset of articles deemed relevant based on the three inclusion criteria described in the published manuscript. In the initial search cycle, we conducted searches in the PsycInfo, Google Scholar, Web of Science, and IEEE Xplore databases.
Our goal was to be as inclusive as possible, so we used the term “Facebook” as well as search terms such as “online social network” and “social network analysis” to identify other articles relating to Facebook that may not have explicitly used the term “Facebook” in the fields searched by keyword searches (e.g., the title, abstract, and list of keywords). This method yielded articles published primarily in peer-reviewed academic journals and conference proceedings, which was the focus of the review. After the initial search was completed, we reviewed the references cited by each relevant article and identified any new reports that were then included in a second round of articles to examine. This two-step approach yielded a wide range of articles relevant to Facebook. With each new relevant article, we browsed the references, searching for additional relevant articles, and continued this iterative process until no new relevant articles were found.
The first exhaustive literature search was completed on January 1st, 2011, and this review served as the basis for the creation of our five theme-based categories. To provide the most thorough review for our readers and increase the usefulness of the bibliography, shortly before publication we updated our bibliography through January 1st, 2012. The follow-up literature search also served as a useful test for our 5-category structure. Given the magnitude and diversity of the article from 2011 (186 articles), the five categories proved to be a rebmarkably effective organizational framework, and the final list based on this procedure is represented in the 'The List By Category' tab. To keep this site as relevant to current researchers as possible, we periodically update the "The List by Date' tab and "The List A-Z" tab to capture the most recent articles from 2012.
Note, we did not restrict our search to the social sciences and any articles that met the three inclusion criteria were included regardless of their disciplinary affiliation. So it is quite possible that our review captures every report written on Facebook, not just those in the social sciences. However, we did not specifically search databases focused on other disciplines such as Law and Management, so we restrict our claims to the social sciences.
Review of Data Collection Methods
The researchers cited in our literature review used three principle methods to collect Facebook data: Recruitment of participants in offline contexts, recruitment of participants via Facebook applications, and data crawling. We hope the following summary of these three approaches will help future researchers decide on the best method for their data collection design. If you have a specific question about setting up a Facebook study, we encourage you to reach out to other researchers through our Facebook Group page
Recruitment of participants in offline contexts. Several studies recruited volunteers (often college students) in offline contexts. For example, to determine the percentage of users verses non-users, one early Facebook study used an offline survey, which was a necessary approach to capture a representative sample (Acquisti & Gross, 2006). This method was especially effective when comparing offline and online behavior or when evaluating users verses non-users.
Recruitment of participants via Facebook applications. Facebook Inc. makes it simple for third party developers to create applications and surveys (see http://developers.facebook.com) and many researchers have taken advantage of this resource (e.g., Vajda, Ivanov, Goldmann, & Ebrahimi, 2011). For example, one research team created three popular applications (Fighter’s Club, Got Love, and Hugged) where users consented to sharing information when they download one of the applications (Nazir, Raza, & Chuah, 2008). Combined, these three applications have over 8 million users, providing researchers with a huge dataset. MyPersonality, another example of a successful application used to collect data, is an online personality survey designed by psychologists. It provides an excellent model of how to construct a useful survey using a Facebook application (Stillwell & Kosinski, 2011). Over 4.5 million users have taken the personality survey on this application, and the researchers have been able to gather a complex and detailed dataset. For example, the researchers released a database of triads where all three friends in the group are described with information on demographics, personality, self-monitoring, IQ, workplace, education, and more (Stillwell & Kosinski, 2011).
MyPersonality demonstrates the vast potential of applications as a tool for recruiting participants on Facebook and studying interpersonal interactions. However, the success of applications is far from guaranteed. Some applications designed to collect data (e.g., youjustgetme) have been much less successful at recruiting participants than others (e.g., MyPersonality, MyType, Stumbl).
Data crawling. Data crawling, which involves gleaning information about users from their profiles without their active participation, provides another effective approach to gathering data from a wide range of users (Gjoka, Kurant, Butts, & Markopoulou, 2011; Kurant, Markopoulou, & Thiran, 2011). Data crawling is possible because researchers can implement algorithms to gather publicly available information from Facebook users (see Gjoka et al., 2011 for an excellent review of data crawling techniques). Note, however, that data crawling has become less informative over time as Facebook Inc. has implemented stricter privacy policies. As of March 2011, Facebook Inc. states that data cannot be collected using automated means (e.g., harvesting bots, robots, spiders, scrapers) without the explicit approval of Facebook Inc. (Facebook, 2011b).
The combination of recruiting participants (offline and through applications) and data crawling gives researchers some effective methods for gathering data on Facebook. In addition, investigators can take advantage of data already collected because several Facebook researchers have made their datasets available to the research community (Ginger, 2008; Nazir, Raza, Gupta, Chuah, & Krishnamurthy, 2009; Stillwell & Kosinski, 2011).
Ethical considerations. Before social scientists collect and examine Facebook data, it is imperative to consider the many ethical obligations inherent in Facebook research and OSN research more generally. One important debate has emerged regarding the appropriate methodological standards for research on Facebook. The heart of this discussion focuses on whether research on Facebook constitutes research with human subjects. Some ethics scholars contend that data mining projects harvest publically available data so they do not meet the regulatory definition of human subjects research, and therefore researchers should not have to gain approval from Institutional Review Boards (IRBs; Schrag, 2010; Solberg, 2010). However, this argument applies only to information that is publically available. Some users do set their privacy settings to allow access to everyone but many users opt for more restrictive privacy settings. If researchers are collecting private information or interacting with Facebook users, then there is an ethical obligation to adequately inform users about the research, gain their consent, and protect their information (Solberg, 2010).
We urge caution and due diligence when conducting research on Facebook. The website Internet Research Ethics (internetresearchethics.org) provides an excellent resource for learning about current IRB standards and how they are being implemented in regards to Facebook. In addition, several articles examine ethical and methodological issues associated with gathering information on Facebook (see Buchanan & Ess, 2011; Mazur, 2010; Solberg, 2010; Zimmer, 2010).