The analysis of social networks has boomed recently, mainly as online social networking systems such as Twitter allow researchers to access these data. However, the research is less and less focused on the fundamental question on the validity of the data and interpretation of the results. For example, Golder et al. (2007) use the word 'friend' in quotes while describing their results.

To enhance the discussion around the validity of results, our work contributes a categorization of social network data. We also discuss the differences of the data sources, especially highlighting the fact that different data sources disclose different kinds of networks. Our approach is to examine social networks based on several sources of data, and thus acquire a richer data set. Based on this extended data set, we are more equipped to understand the social relations represented via links between nodes.

After reviewing the existing literature, we make two observations of social relationships in online services. Firstly, the friendship data may be shared in public or with the specific group of users of that service - this may affect how people perceive and use these relationships, especially when compared with the private displays of relations (e.g., Donath & boyd, 2004). On the other hand, people interact only with part of their social relations (e.g., Golder et al., 2007) and research has started to focus from statical networks to more dynamical activity based networks (e.g., Huberman et al., 2009).

Based on the existing literature, shortly discussed above, a 2x2 matrix can be developed. Relations may be public or private and active or passive. For instance, those relations with which you use Instant Messaging can be considered private and active whereas Facebook friends are passive and public. As they are different in this nature, also the conclusions based on the analysis should differ.

After confirming that the data measure the phenomenon desired, one should use several kinds of data sources to really understand the social structures behind the group under study. We claim that multiple data sets should be used when measuring social relations. McPherson et al. (2001) have also concluded that the priority for future social network researchers should be to gather dynamic data on multiple social relations. By studying existing research and our own empirical data (e.g., Karikoski & Nelimarkka, 2011), we discuss the opportunities and challenges of using multiple data sets to cover the same group.

Abstract references

Donath, J. and Boyd, D., 2004. Public displays of connection. BT Technology Journal, 22 (4), pp.71-82.

Golder, S.A., Wilkinson, D. and Huberman, B.A., 2007. Rhythms of social interaction: messaging within a massive online network. In: The 3rd International Conference on Communities and Technologies (C&T2007). East Lansing, Michigan, USA 28-30 June 2007.

Huberman, B., Romero, D.M. and Wu, F., 2009. Social networks that matter: Twitter under microscope. First Monday, 14 (1-5).

Karikoski, J. and Nelimarkka, M., 2011. Measuring social relations with multiple data sets. International Journal of Social Computing and Cyber-Physical Systems (IJSCCPS), 1 (1), pp.98-113.

McPherson, M., Smith-Lovin, L. and Cook, J.M., 2001. Birds of a Feather: Homophily in Social Networks. Annual Review of Sociology, 27, pp.415-444.