This week I spent Monday and Tuesday at the Connected Pasts workshop and meeting in London, learning about network and complexity science and its application to archaeology and the past. The two days consisted of a three hour workshop introducing and exploring one of a number software packages which can be used to analyse networks on Monday morning, followed by talks and key note speeches on Monday afternoon and through Tuesday (full program and abstracts here). The conference was held at Imperial College London, and organised locally by Tim Evans of Imperial, with Tom Brughmans of Konstanz from the wider advisory committee.
The conference was billed as multidisciplinary (within the scope of the past), and largely lived up to that; I chatted to computer scientists, geographers, statisticians, classicists and lots of physicists and archaeologists. It also attracted an international group of speakers and attendees, largely from Central and Western Europe, though with some North Americans as well. It also attracted a predominantly young audience (well, for archaeology at least – but that’s a separate post!), which appeared to be dominated by PhD students and post-docs. All of these factors lead to a really engaging, open group of people which made the conference very enjoyable. Unlike some of the bigger archaeology conferences, people were very interested to talk to everyone, and mixed freely, which lead to some very interesting lunch and coffee conversations.
The workshop was excellent; as you can see from my awkward image above we were coaxed through an introductory tutorial on Cytoscape, one of the free to use software packages available for network analysis, until we were able to produce our own network visualisations from a set of Tom’s data. Getting practical experience manipulating the visualisations and generating summary statistics on the networks was very engaging, and immediately filled my head with excited thoughts about what data I had which might be examined using these techniques.
However a number of the papers later in the first day were quite challenging in their use of archaeological datasets. In particular one of the papers used a network which was based on the transmission of ‘ideas, iconographies, models’ between urban centres as a way of generating archaeologically meaningful interpretations of ‘cultural transmission’. Now, as an archaeologist this immediately makes me uncomfortable as the whole idea of diffusion/acculturation/creolisation/’romanisation’ or to put it simply, the way people and cultures interact and change each other, is a really big debate. It isn’t as simple as ‘x culture adopts y culture’, there’s always transfer both ways and sometimes new cultures occur which attach entirely new meanings to the objects we might associated with x or y.
Consequently the idea that you could just say ‘this idea moved from this town to this town in this direction’ seems a substantial simplification of a complex archaeological concept, and at the least needs heavy contextualisation and an established theoretical basis. For example, if the idea is ‘lets build hypocausts’ how do you say what a hypocaust really means? The appearance of hypocausts might mean the appearance of Romans, a fashion for Roman things, a change in the climate, or the appearance of a hypocaust builder to name just a few. I suspect that contextualisation may have been in the original work, though not in the paper as presented due to the network science focus, but it’d be interesting to see what the justification behind the data choices was.
Thinking about Roman/non-Roman contact made me a little concerned about using network analysis for dealing with cultural change after conquest or contact. From what I learned at the workshop, network analysis is built around nodes (e.g. buildings, sites, regions) and arcs between them (e.g. roads, movement of goods), and the fundamental unit is a pair of nodes. This emphasis on pairs and a directional relationship between them (at least in most of the papers on this I saw) seems to lead to a natural emphasis on polarity and unidirectional cultural transfer, which are ideas we’re trying to drag ourselves away from in the archaeology of the Roman period.
I wonder if non-directional links between nodes might alleviate some of the misgivings I have with the network analysis of ‘cultural’ contact between cities or states. A non-directional link would create space for the interaction between the two cultures to involve more than just ‘x culture adopts this from y culture’ inherent in the directional links and which seems like a simplification too far, leaving out much of the rich human context we’re trying to get to. The underlying archaeological theory indicates that when cultures interact the communication and adoption goes both ways even when one culture has substantial power over another, so it would make sense to give this space within the network analysis.I don’t know how this would change the outcomes, but it would stop the analysis technique from clashing at a fundamental paradigmatic level with much of the archaeological theory around cultural interaction in the Roman period at least. I need to find some network scientists I tie down long enough to explain all this to, and get their opinions!
Hopefully many of the slides from the two days will be up on figshare at some point, and I will put together a Storify of the tweets from the conference when I have recovered from the 5:45am mornings! Beyond those above, there were a number of other really interesting points raised at the meeting which hopefully I will put together into a conference review soon, probably for PIA or possibly The Crucible (HMSnews), or even a few more blog posts. But the one thing that I took home from the conference was that network and complexity analysis is definitely happening in archaeology, and it may just become a very important part of the analytical toolbox.