Review of ORBIS

Quantitative methods matter in the digital humanities, and are coming to play an ever more prominent role in their discourse. This is driven, in part, by a great increase in interest in simulation modelling. Often borrowing methods and software from social scientists, who use them to interrogate vast datasets on populations, historians and archaeologists are coming to use quantitative tools to model historic scenarios. The use of simulation modelling in history and archaeology is still regarded as something of a ‘dark art’ in many quarters. After all, the purpose of historical or archaeological enquiry is to reconstruct the past from evidence, be that evidence textual or material. Simulation, on the other hand, begins with evidential parameters and builds ‘could-be’ hypotheses, which are in some way credible, upon them. Such hypotheses cannot be tested or reproduced from other data, and as such — as methods of understanding the past — they are usually treated as a means to the end of interpretation, not an end in themselves.

One new project experimenting in this area is ORBIS, designed by Walther Scheidel, a Roman historian, and developed by the technology team of Elijah Meeks, Karl Grossner, and Noemi Alvarez at Stanford University. Put simply, the aim of ORBIS is to model the costs and times of travel between different points within the Roman Empire, over land or by sea or river. It does this by identifying factors such as wave height, terrain, mode of transport (e.g. camel train versus rapid military march) and mean climatic conditions. Using these, it computes how long it would take different categories of traveller using different means of transport, and at what cost, to traverse those distances by the optimal route. Results can be filtered by the fastest, cheapest, and shortest options. The site is free to use and, it should be noted, is a work in progress, with numerous upgrades, both technical and conceptual, promised for the future. Therefore, the observations that follow should be read not as criticisms, but as commentary on possible directions for its future development.

The site is structured with a simple and easy to navigate layer of content. Different tabs present an introduction and sections on ‘Understanding’, ‘Building’, ‘Using’, ‘Mapping’, and ‘Applying’ ORBIS, with various sub-tabs demonstrating particular aspects or works in progress on visualisation. The text and images of these are also available as a PDF download, which somehow gives credence to the authors’ claim that the site is not just a site, but also an online scholarly publication incorporating and building a software tool.

Usability aside, ORBIS is a new foray into rather perilous territory. The stated aim is ‘to understand the dynamics of the Roman imperial system as a whole.’ It explicitly eschews lower levels of detail and granularity, seeking rather to reconstruct, or rather model, systemic patterns of travel from pre-determined parameters. In its scope especially, ORBIS is distinct from many previous experiments in historical simulation, such as the Road to Manzikert project,[1] or the Battle of Trafalgar.[2] These projects (which are actually Agent-based Modelling exercises, and although ORBIS is not such a project it nonetheless identifies ABM as a possible future direction of expansion) focus their simulations on distinct events, such as the progress of the Byzantine army to Constantinople in AD 1071, or the naval engagement of 1805; or on processes which are tightly constrained in time or space, with logical and easily-justified boundaries. ORBIS, with a set of environmental parameters across at least two climate zones, travel scenarios spanning three modern-day continents and some twenty modern-day national entities, and with few sharp lines around its edges in a qualitative historical sense, is therefore an undertaking of admirable, even daunting, ambition.

In any historical simulation project, the most important thing is to be clear about the limitations. In this, ORBIS cannot be faulted. The ‘Building ORBIS’ tab presents a long and detailed account of the principles guiding the selection of parameters for the modelling, and the evidence from which the quantitative aspects of the modelling are derived. This includes salient observations such as the fact that the average time computed for a ship’s journey between any two ports will likely underestimate any actual time, since any ship is susceptible to a range of factors that might slow it down such as damage, piracy, or poor seamanship; whereas there are far fewer factors which could result in a sea journey being completed faster than would otherwise be anticipated by any predictable configuration (ancient or modern).

This set of quantitative predictions is constrained throughout by a focus on systemic features. The model includes only main arterial roads, rather than minor tracks, and excludes waterways that could have been navigated by small vessels or rafts. ORBIS keeps carefully to its turf, and avoids the temptation to be an oracle for any scholar wanting to construct any journey at any scale. It isn’t, and it says it isn’t. On the contrary, the scholarly referencing of ORBIS is impeccable, and the inclusion of well over 150 bibliographic references further supports the authors’ claim that this is a research tool and a scholarly publication combined.

The quantitative model that ORBIS relies on is explained in detail in the ‘Building ORBIS’ section, and its limitations carefully outlined. Secondary studies such as Pascal Arnaud’s analysis of shipping are drawn on, as are pieces of primary evidence such as the tetrarchic price edict of AD 301, which is used to calculate the cost of maritime freight. These are used to justify the model quantitatively. The 751 locations are derived from the standard reference work for the ancient world’s geography, the Barrington Atlas. A total of 84,631 kilometres of the Roman Imperial road network is mapped. The coastal sea-lanes mentioned in the historical sources are privileged in the ORBIS model. It is interesting that in a previous iteration of the model, the project considered using a minimum range approach, whereby connections between coastal sites within a minimum radius (of 500 miles) would have been privileged — the dismissal of a purely abstract quantitative approach in favour of one derived from sources is, I think, an interesting reflection of how the humanities’ needs for quantitative methods differs from those of the social sciences.

I look forward to the promised future upgrade where the user will be able to view information on seasonal hazards into the outcomes of the calculations they have submitted. River transportation is, in general, far more difficult to reconstruct, due to the shifting character of the landscape, modern human intervention in Europe’s waterways, and paucity of documentation. Accordingly, ORBIS’s model relies on comparanda from the medieval period. This of course introduces additional levels of complexity and uncertainty, and one area that must be developed further is the question of how this greater cumulative fuzziness might affect individual applications of the model.

So: the constraints of the model are described in detail, and ORBIS is scrupulously careful not to make unfounded claims. However, the constraints themselves occupy a place in the debate around quantitative methods in the reconstruction of the past, and future iterations of the project must — in my own view — continue to exercise caution against the temptation to lapse into environmental determinism. Future iterations must also continue to make clear that ORBIS does not try to simulate agency, and that this requires the user to use it only in the context of their own analysis or interpretation, and not as an independent generator of research conclusions. For example, ORBIS states that ‘while much of the sea is in theory navigable without major restrictions, sailors would often follow established routes, and certain roads and rivers were more heavily used than others and hence more vital to the functioning of the system.’[3] This seems to assume that the only constraints followed by sailors navigating water was wind and currents, and ‘black boxes’ the reasons why certain routes became ‘established’ and, crucially, how these factors might change over time. One can — as I do — fully accept ORBIS’s premise that ‘No one traveler would encounter such [simulated] outcomes except by chance,’ and appreciate that the processes being simulated are pan-Imperial and systemic.[4] However, when one considers the great events that shaped the Roman world over time — wars, assassinations, expansions, incursions — the limitations of adopting a purely abstract approach based on systemic parameters, become even more apparent. This tantalizes the user with the possibilities that, one hopes, future iterations of the tool will explore. For example, it is difficult to see how the processes we can discern from non-simulative historical sources, such as the consolidation of the Roman Empire’s frontiers between AD 140 and AD 160 could be explored using a tool whose only constraints are environmental. One could speculate that this is merely one factor of a similarly high/systemic level, specific to the Roman period, which the user could include. As the project develops, the ‘Applying ORBIS’ tab, which presents instances of the model’s use in academic work, must expand and document the tool’s use in these areas.

My own experiment with the ‘Mapping ORBIS’ section of the site, where one can try the tool, concerned a relatively short-haul route in northern Britannia between Segedunum (Wallsend, Newcastle) and Luguvalium (Carlisle) — about five sixths of the 74 mile expanse of Hadrian’s Wall — mainly because I have walked this route myself (or at least the sections of it followed by the modern-day National Trail). ORBIS computes the ‘cheapest’ option as an entirely plausible 4.7 days. However, this includes river transportation as far as Corstopitum (Corbridge), and whether the Tyne was navigable that far upriver in this period is a far from straightforward question, underlining the uncertainties raised about ancient river conditions. When one seeks to exclude river transportation from the calculation, an error message is returned. It seems not to be possible to specify travel overland from Segedunum to, for example, Cilurnum on the Wall itself, and then link up with the overland route from thence to Luguvalium. While one takes the point that a section of river travel is needed for the journey to be literally fastest, cheapest, or shortest, it is surely important for the user to be able to include only modes of transport that they know they are interested in. Otherwise it is the model, not they, who is in the driver’s seat. The experience itself however is user-friendly and straightforward, and one can easily see how it could be usefully combined with the ‘Network Visualisation’ tool, which allows you to see and navigate the connections between the different points, under the ‘Understanding ORBIS’ tab.

It is a personal belief of mine that the concepts of ‘quantitative’ and ‘qualitative’ data are direct imports from the social sciences, and we in the humanities do not always think in enough detail about what this distinction means for humanities material. ORBIS is an exciting and innovative experiment in simulation modelling in history and archaeology, which starts to critique this distinction; even if, by its own clear admission, it does not yet have the answers. This project is only just starting out, and it has made a convincing pitch for its approach to what we might think of as a qualified quantitative network and spatial visualisation of a generalised view of travel in the Roman world. The possibilities for the future are immense, and one can only begin to imagine the strides that could be made were it to follow up on the aim stated in the documentation to build links with the Pleiades and Pelagios projects. I, and I am sure many others, will be watching ORBIS’s progress with great interest.

 

  1. [1]Vince Gaffney, Georgios Theodoropoulos, Rob Minson, and Phil Murgatroyd, “The Road to Manzikert,” Digital Classicist, 2009. http://www.digitalclassicist.org/wip/wip2009-02pm.pdf.
  2. [2]Giuseppe Trautteur and Raniero Virgilio, “An agent-based computational model for the Battle of Trafalgar: a comparison between analytical and simulative methods of research,” Proceedings of the Twelfth IEEE International Workshops on Enabling Technologies: Infrastructure for Collaborative Enterprises, 2003. http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=1231440.
  3. [3]Understanding ORBIS tab.
  4. [4]Ibid.

About Stuart Dunn

Stuart Dunn is a Lecturer in the Department of Digital Humanities at King's College London, where he was previously a Research Associate and Research Fellow. He teaches digital preservation, cultural heritage, and archaeology and material culture. Stuart graduated from the University of Durham with a PhD in Aegean Bronze Age Archaeology in 2002, conducting fieldwork and research visits to Melos, Crete, and Santorini. His interests include ancient geography, the representation and enforcement of boundaries and frontiers, and the representation of human motion in 3D digital reconstruction of cultural heritage, and in museums. Most recently he has worked on the virtual reconstruction of social space in Southern British Iron Age round houses, and on the geography and digital representation of historic English place names. He is also interested in the theory and practice of digital communities using social media, and the development of digital infrastructures to support collaborative research.