As I mentioned before, I recently published a review of formal network methods in archaeology in Archaeological Review from Cambridge. I want to share the key problems I raise in this review here on my blog, because in many ways they are the outcomes of working with networks as an archaeologist the last six years. And yes, I encountered more problems than I was able to solve, which is a good thing because I do not want to be bored the next few years 🙂 In a series of four blog posts I draw on this review to introduce four groups of problems that archaeologists are faced with when using networks: method, data, space process. The full paper can be found on Academia. This first blog post in the series discusses methodological issues, enjoy 🙂
Like any other formal techniques in the archaeologist’s toolbox (e.g. GIS, radiocarbon dating, statistics), formal network techniques are methodological tools that work according to a set of known rules (the algorithms underlying them). These allow the analyst to answer certain questions (the network structural results of the algorithms), and have clear limitations (what the algorithms are not designed to answer). This means that their formal use is fundamentally limited by what they are designed to do, and that they can only be critically applied in an archaeological context when serving this particular purpose. In most cases, however, these formal network results are not the aim of one’s research; archaeologists do not use network methods just because they can. Instead one thinks through a networks perspective about the past interactions and systems one is actually interested in. This reveals an epistemological issue that all archaeological tools struggle with: there is a danger that formal networks are equated with the past networks we are trying to understand (Isaksen 2013; Knox et al. 2006; Riles 2001). In other cases, however, formal analysis is avoided altogether and concepts adopted from formal network methods are used to describe hypothetical past structures or processes (e.g. Malkin 2011). Although this sort of network thinking can lead to innovative hypotheses, it is not formal network analysis (see reviews of Malkin (2011) by Ruffini (2012) and Brughmans (2013)). However, such concepts adopted from formal network methods often have a very specific meaning to network analysts and are associated with data requirements in order to express them. Most crucially, when the concepts one uses to explain a hypothesis cannot be demonstrated through data (not even hypothetically through simulation), there is a real danger that these concepts become devalued since they are not more probable than any other hypotheses. Moreover, the interpretation of past social systems runs the risk of becoming mechanised when researchers adopt the typical interpretation of network concepts from the SNA or physics literature without validating their use with archaeological data or without modifying their interpretation to a particular archaeological research context. This criticism is addressed at the adoption of formal network concepts only. It should be clear that other theoretical concepts could well use a similar vocabulary whilst not sharing the same purpose or data requirements, in which case I would argue to refrain from using the same word to refer to different concepts or explicitly address the difference between these concepts in order to avoid confusion.
Although it is easy to claim that the rules underlying formal network techniques are known, it is less straightforward to assume that the traditional education of archaeologists allows them to decipher these algorithms. Archaeologists are not always sufficiently equipped to critique the mathematical underpinnings of network techniques, let alone to develop novel techniques tailor-made to address an archaeological question. For many archaeologists this means a real barrier or at least a very steep learning curve. Sadly, it also does not suffice to focus one’s efforts on the most common techniques or on learning graph theory. Like GIS, network analysis is not a single homogeneous method: it incorporates every formal technique that visualises or analyses the interactions between nodes (either hypothetical or observed), and it is only the particular nature of the network as a data type that holds these techniques together (Brandes et al. 2013). For this purpose it draws on graph theory, statistical and probability theory, algebraic models, but also agent-based modelling and GIS.
A thorough understanding of the technical underpinnings of particular network techniques is not an option; it is a prerequisite for a critical interpretation of the results. A good example of this is network visualisation. Many archaeologists consider the visualisation of networks as graphs a useful exploratory technique to understand the nature of their data, in particular when combined with geographical visualisations (e.g. Golitko et al. 2012). However, there are many different graph layout algorithms, and all of them are designed for a particular purpose: to communicate a certain structural feature most efficiently (Conway 2012; Freeman 2005). These days, user-friendly network analysis software is freely available and most of it includes a limited set of layouts, often not offering the option of modifying the impact of variables in the layout algorithms. Not understanding the underlying ‘graph drawing aesthetics’ or limiting one’s exploration to a single layout will result in routinized interpretations focusing on a limited set of the network’s structural features.
Archaeologists who consider the application of network methods to achieve their research aims must be able to identify and evaluate such issues. Multi-disciplinary engagement or even collaboration significantly aids this evaluation process.
Brandes, U., Robins, G., McCranie, A., & Wasserman, S. (2013). What is network science? Network Science, 1(01), 1–15. doi:10.1017/nws.2013.2
Brughmans, T. (2013). Review of I. Malkin 2011. A Small Greek World. Networks in the Ancient Mediterranean. The Classical Review, 63(01), 146–148. doi:10.1017/S0009840X12002776
Conway, S. (2012). A Cautionary Note on Data Inputs and Visual Outputs in Social Network Analysis. British Journal of Management. doi:10.1111/j.1467-8551.2012.00835.x
Freeman, L. C. (2005). Graphic techniques for exploring social network data. In P. J. Carrington, J. Scott, & S. Wasserman (Eds.), Models and methods in social network analysis (Vol. 5, pp. 248–268). Cambridge: Cambridge University Press. doi:10.3917/enje.005.0059
Golitko, M., Meierhoff, J., Feinman, G. M., & Williams, P. R. (2012). Complexities of collapse : the evidence of Maya obsidian as revealed by social network graphical analysis. Antiquity, 86, 507–523.
Isaksen, L. (2013). “O What A Tangled Web We Weave” – Towards a Practice That Does Not Deceive. In C. Knappett (Ed.), Network analysis in archaeology. New approaches to regional interaction (pp. 43–70). Oxford: Oxford University Press.
Knox, H., Savage, M., & Harvey, P. (2006). Social networks and the study of relations: networks as method, metaphor and form. Economy and Society, 35(1), 113–140. doi:10.1080/03085140500465899
Malkin, I. (2011). A small Greek world: networks in the Ancient Mediterranean. Oxford – New York: Oxford University Press.
Riles, A. (2001). The Network inside Out. Ann Arbor, MI: University of Michigan Press.
Ruffini, G. (2012). Review of Malkin, I. 2011 A Small Greek World: Networks in the Ancient Mediterranean. American Historical Review, 1643–1644.