Agent Modeling and the settlement of the Ottawa Valley

I’m working on a paper at the moment, for a special issue of the Journal of Canadian Studies. Provisional title is ‘The Digital ‘What If…’: The Potential for Agent Based Modeling in History Computing’. I’m taking my Travellersim agent-based model (which I’ve used for studying the emergence of settlement patterns in the proto-historic Tiber Valley in Italy), and running it on the Upper Ottawa Valley.

“What!” I hear you say, “you can’t be proposing that the situation in Iron Age Italy is the same as 19th Century Canada!”.

I’m not actually proposing that. In the Travellersim model, travel is envisioned as so much distance per day, and that travellers travel to places that are attractive to them (for whatever reason). And how is attractiveness measured? It depends on how many other travellers have already visited that place (making a feedback loop). Travellersim generates likely patterns of interconnections between settlements in a region, focusing on “situation”, the human patterning of social connections over space, rather than “site”, the physical location of the settlements themselves. So long as the rules are framed at the level of an individual agent, I argue that the simulation may be applied to different times and places: as long as they aren’t travelling by rail, airplane, or automobile. The model emphasises the role of individual social interactions for the emergence of larger cultural regions distinct from physical geography.

This is particularly interesting when you consider the artificial boundaries of the Ottawa Valley: the division into Ontario and Quebec. Since about the 1960s, the Quebec provincial government has taken steps to try and curb cultural interaction with our neighbours in Ontario (full disclosure: I’m an Anglo Quebecer living in the Quebec side of the Ottawa Valley). For instance, minor hockey teams in Shawville (Que) are not allowed to play against teams in Renfrew (Ont) or Pembroke (Ont), but are forced (via the blunt instrument of insurance policies!) to travel to Gatineau instead. But when I run this model on pre-railway settlement data, it suggests (amongst other things) cultural connections across the river at particular points: BristolArnprior, Campbell’s Bay/Calumet Island – Beachburg, Allumettes Island/Chapeau-Pembroke.

This makes sense. Families span both sides of the river; my grandfather (who lived in Bristol), used to cross the ice in the winter time to do his shopping and so on in Arnprior. Bricks in the farmhouses in Bristol Township tend to come from Arnprior brickyards (when this can be determined), than from the Shawville brickyards (in the next township to the west, Clarendon).

So in this paper, I’m looking at

  • what patterns does the model suggest;
  • is there historical support for these patterns, or are they artefacts of the model (if they are artefacts, I’ll have to determine what’s going on, and why, and see if the model can be salvaged!)
  • assuming that there is historical support, I’ll then explore how these emergent cultural spaces map against political boundaries
  • and finish up by seeing if there are historical artefacts that can be explained with reference to the conflict between the cultural and political spaces.

A tall order, non?

I’ve begun with Bristol Township. The data I’m using comes from a heritage survey I did in 2003; 46 points based on:
-original homesteads
-mill sites
-churches -> on land donated by citizens, rather than located by Church organisations to take advantage of population
Given this landscape, what sort of territories emerge? Do they correspond with ‘zones’ that make sense in later history? Which areas become ‘centres’, and do they correspond with actual village/town development?

Picture of start of model below (Bristol township is not square shaped, but more triangular, tapering to an almost point in the north):

And here is the model after a few time-ticks:

The visual output is not the actual data that gets analysed; it’s more of a short-hand for representing which sites were the most attractive to travellers. I use the tools of social network analysis to evaluate the underlying networks between sites, and between travellers, to work out the actual ‘zones’ (see my paper in CAA 2006 for all the nitty gritty).

Anyway, the paper isn’t due until December; today just felt like a day to get working on it. I am, as ever, interested in what people think about this work…