Friday, October 26, 2007

Follow-up on discussion, week 10

Great discussion on Thursday. There was definitely a lot more fodder for discussion than we could get into in just that class period.

A few notes:
Fred mentioned, in regard to the example about schooling behavior in fish, that just because you show that the emergent phenomena could result from localized and 'blind' interactions doesn't mean that you actually have shown it. While I think the school thing is still a good example of an emergent phenomenon, her comment points directly to one of the biggest issues in the use and interpretation of agent based models. You can give agents rules and tweak parameters until you get all kinds of different patterns. Once you've demonstrated that your model can generate a pattern like the one you're interested in, can you conclude that you've therefore explained it? (there is a tendency for modelers in this area to act as if generating the 'right' pattern with a set of rules is the same as explaining it). Not only could lots of different types of localized interactions, based on different rules and parameter values, potentially generate the same pattern but other external or top down controls might still be relevant even though the model generates a pattern without them. This creates a difficult situation for the use of such models but a lot of applications have shown that you can make more refined empirical predictions, which can in turn be tested with additional observation or fieldwork. Lansing's work may be the very best example of this in the social sciences. Keep in mind that the paper we read was from 1993 and he has done a lot of stuff since then. Check out his website if you're interested.

Als0 relevant to the Lansing and Kremer paper is the issue of agency. We all seemed to take some kind of issue with the contrast between 'blind' and 'deliberate' selection and what roll human agency played in Bali temple networks. I'm not sure we resolved this in class but we came up with at least two types of possible interpretation: 1) we disagree with Lansing and Kremer and think that they have shown the opposite of what they say they've shown - that blind localized interactions generate the temple network and that you don't need to invoke some special human agency at all - but that depends on what you think agency is. 2) that if you 'read between the lines' of their paper they are actually disagreeing with how a lot of anthropologists use the concept of agency and they are just doing so in a careful and stealthy way. It is most likely the case that they are making the point that the agricultural system falls apart without conscious human action, not at the level of the whole system but in terms of intentionally maintaining and managing plots of land, whereas other 'natural' systems presumably don't require this explicit planning. This would assume that human planning is qualitatively different from the planning done by say, a beaver. Agency can be a bit of a slippery concept. Regardless of these issues of interpretation, that this complex social structure could be generated as a 'self-organized' process of neighbor-to-neighbor communication is hugely important, provocative, and definitely not a typical variety of anthropological explanation.

I also liked how we kept finding more linkages between the papers the more we talked about them. Perhaps in the early days before the establishment of the temple networks and the social structures they help maintain, there would have been more problems of synchronization as the system was in more of an 'r' rather than a 'K' phase, as Paul pointed out. Maybe shrimp aquaculture is not so well-organized or structured and top-down controls are necessary to keep it from going chaotically out of control? We could attempt to categorize these issues and others with respect to Holling's phases to see if we are able to gain insight via application.
Have a great weekend. We'll be posting more shortly...

No comments:

Locations of visitors to this page