Dominik Froehlich, Marc Sarazin “Carelessness in system specification is probably a more serious issue for network analysis than for much survey analysis” - this quote by Laumann et al. (1983, p. 19) is becoming more and more relevant, given the increasing popularity of social network analysis (SNA) in a variety of academic disciplines. Criticism has been voiced against the validity of many SNA studies. This includes, for instance, the missing generalizability of much SNA work, theoretical underspecification, or conceptual slippage of the term “network”.
Lots has been written about the validity and reliability of different social network measurement techniques and methods, but far less about the validity of the notion of 'network'. In this presentation, we pick up this gap and focus on a specific instance of validity that we call network validity. Some thoughts have been spent on the topic, especially under the banner of the boundary specification problem. However, with this framework of network validity presented here, we hope to show that there are more considerations to be taken into account to have a valid network study. The main question we pose is how to determine whether the thing that we’re studying using network methods actually constitutes a network and what kind of network it is. In developing our argument, we refer to Dellinger and Leech’S (2007) mixed methods validation framework (given that SNA can be considered an inherently mixed method; Crossley & Edwards, 2016).