Comparative analysis of articulated and behavioural social networks in a social news sharing website

Abstract

This study analyses and contrasts the explicit (articulated) and implicit (behavioural) social networks on the Spanish Digg-like social news website meneame.net. The explicit network is given in the form of declared but not necessarily bidirectional friendship links; the behavioural network is extracted from conversations through comments to the shared links. These two directed social networks and their intersection are analysed and described in detail, which leads to some important conclusions about user behaviour on link sharing websites and online conversation habits in general. We find that reply interactions are more likely to occur between non-friends and that these interactions are (if bidirectional) also more balanced in the case of non-friends. A k-core decomposition of the networks reveals a fundamental difference in the practice of establishing behavioural and articulated links.

Keywords:

• Social media
• Social network analysis
• Link sharing
• k-Core decomposition

Acknowledgements

The authors wish to thank Vicenç Gómez and David Laniado for helping in the evaluation of this study and Meneame.net for allowing access to an anonymised dump of their database. This work has been developed inside the i3media project (CDTI 2007–2010) and partially funded by MediaPro, the Centro para el Desarrollo Tecnológico Industrial (CDTI) within the Ingenio 2010 initiative. Yana Volkovich acknowledges support from the Torres Quevedo Program from the Spanish Ministry of Science and Innovation, co-funded by the European Social Fund.

Notes

1. http://www.zephoria.org/thoughts/archives/2009/07/28/would_the_real.html (accessed 22 April2010).

2. http://www.meneame.net

3. See http://www.blog.meneame.net/2009/05/07/explicacion-simple-del-algoritmo-de-promocion-de-noticias-promote/ (accessed 22 April 2010) for details.

4. Described in http://www.meneame.wikispaces.com/Karma (accessed 22 April 2010).

5. If the network is directed, then one can use either the in- or the out-degree of the node v. There exist therefore two different k-core decompositions for directed networks. Such an extension of the original definition of Seidman (1983) has been first proposed by Batagelj and Zaversnik (2002).

6. Note that, as can be seen from the top row of Table 4 , some of these users do not have any links in the reduced networks, since all of their link targets are users who do not fulfil the above condition.

7. Since the measure is symmetric we calculate the statistics only for positive values of ρ _reply and use only half of the cases when ρ _reply =0.

8. Another possibility to explain missing back-replies between friends would be the existence of other communication channels (besides those of Meneame) between friends to interact. However, given the nature of public discussion it seems to be unlikely that the discussion switches exclusively to such hidden channels if a friend's reply would have had negative connotations.