Tuesday, May 20, 2008

Using "Facebook" for Research

This had led me down a rabbit's hole...

Apparently, a group of French researchers have been going through medieval land transactions in the province of Lot (in S. France) to reconstruct social networks. The title of this piece is "Social Networking gets Medieval." From the title and my short description, I was thinking that they had constructed a computer model (a la Facebook) to reconstruct social networks in late medieval France. But that's not quite what these researchers did.

Publishing their findings in the journal Neurocomputing, these researchers went through legal transactions to figure out who knew who and where and when. Their conclusion? There are lots of pretty graphs and they figure out that these statistical models can help historians in some ways, although social context matters greatly and a "perfect community" (whatever that means) can't really be constructed.

I don't really know what to make of all this. On the one hand, computer modeling of intricate relationships - especially if it were user-friendly and widely available on the web - would be tremendously useful, especially if you're dealing with (what seem to be) tight social networks, like we do indeed have in most areas of the Middle Ages (and elsewhere). One could log-in, add a name and some information about that person, and have the program automatically draw out possible connections to other people that other scholars have found.

On the other hand, I think what they're doing is just called prosopography and I think Jonathan Jarrett (among others) knows a heck of a lot about. These French mathematicians are, in a sense, reinventing the wheel here, rather than utilizing the voluminous literature on social networks in the Middle Ages that generations of (predominantly German) scholars have put together. Look at what Paul Ormerod says about quantitative analysis. He's just plain wrong. This model is all numbers without the analysis.

Now, my head's a little fuzzy. When I saw Nature.com's headine, I was expecting Facebook to be made relevant to me as a scholar. Now, I'm dealing with free-lance economist/ historian consultants and statistical analysis of social networks. Help!

10 comments:

tenthmedieval said...

You have figured me right, here, Matt. This story gives me the risen hackles, just because not just I, but thirty years previous of scholars have been doing this for ages. Barbara Rosenwein's work on Cluny is the real deal, for example, using what was then cutting-edge tech (1985) to do group-search analysis on a database of thousands and thousands of names to produce associations. The only way in which this is not that is that whereas she was thinking conceptually of asociations, they are thinking of networks. And there I would like to claim prior art, if only I hadn't got it from Wendy Davies... Pfah. Which is not to say that it's not valid, I just wish Journal of Neurocomputing had asked a historian to review...

tenthmedieval said...

Something of a knee-jerk reactiion there, now that I look deeper. I think there is originality here, and that that lies in the presentation of the data and the algorithms used to analyse it. What I won't know till I read the paper, and possibly not after that given it seems heavily technical, is whether doing it that way tells us anything that the old-style didn't. If not, it would be a parallel to that project I posted about a while ago using DNA mutation as an interpretation aid to scribal error in copying manuscripts; very shiny, but at best a way to tell us again what we already knew.

Another Damned Medievalist said...

Prosopography? Land transactions? Putting together networks and relationships? Whatever will they think of next? Is Régine LeJan unknown in France? (Actually, I think she could be Belgian, but she bloody works in France, or did, the last time I looked, which was last summer when I ploughed through most of her stuff...).

Nathalie Villa said...

I am one of the authors of the article published in Neurocomputing. I have to say that the article does not intend to provide new informations about medieval society but to show how data mining tools can help to analyze the structure of a big network. Historian work that can be deduced from this will come probably next year but.
On an historical point of view, the size of the corpus that we have recorded is very interesting (in France, at least) as several thousands of people that were living during 200 years in a small region were mentionned.
More informations, if you are interested in, by contacting me...
Best regards,
Nathalie Villa

Matthew Gabriele said...

Nathalie, first of all, welcome and thanks for visiting!

I certainly didn't mean to imply that your group, in the article, was specifically interested in doing that sort of analysis. I just wonder how much consultation your group did with historians to see if the research would actually be useful. Moreover, how much did you base your data mining rubric on the mountains of work that's already been done on (especially) medieval social networks?

I have no doubt that such tools could be a goldmine for scholars of the past but I wonder how much work you're replicating unnecessarily.

Nathalie Villa said...

We've got frequent meetings with historians who are a entire part of the project. See, e.g. Florent Hautefeuille http://w3.terrae.univ-tlse2.fr/spip/spip.php?article35 . He is interesting in medieval sociaty until years and probably have references about this subject that I don't have (obviously, I'm a statistician). The interest of our work on his point of view - as long as I can understand it - is :
*) having a way to understand the structure of this large network which deals with people that are fewly mentionned in the classic literature: the people in our network are _all_ peasants and none of them are nobles
*) having tools to identify major families of peasants (that can be usefull to provide deeper research on them)
*) having tools to see how is evolving the network during time. We already see that the structure of the network is globally preserved after the Hundred Year's War (and is very similar to those of modern social network) but that the major families of serfs are changing.

Once again, I am not an historian and do not intend to say interesting historian things: that's the role of historians. But by multi-disciplinary work we can help improve the possibility of understanding a large corpus of documents of historians. Sociologists do that since years and I am really surprised to see that historians can be so suspicious about it.

Finally, to explain why we didn't mimic past works, I have to say that the historians are especially interesting in that corpus of documents and I understood that they are because this corpus is only made of _agrarian contract_ so doesn't tell about the life of the nobles but of the "serfs" and other peasants. Knowing the life of peasants and their social relationship at that time is not so easy (as far as I know) because they are only fewly mentionned in classical litterature...

To conclude, I don't think that the historians that have asked us to associate with them for that work are intending to "replicate work unnecessarily"... But it is certainly not in Neurocomputing that we can deal with it !

tenthmedieval said...

There are of course methodological problems with being sure that someone is a peasant, but I have to admit that them being involved in a contract to till the soil is a pretty good indicator. (Who are the other parties, though?) I'm not going to comment further on this till I've read the Neurocomputing paper, however, and possibly consulted contacts who will understand it better than me. If this does turn out to be useful, I want to make sure I can use it :-)

tenthmedieval said...

It's taken long enough, but I have now given this a proper go over at A Corner of Tenth-Century Europe.

tenthmedieval said...

Ye saga continueth: another of the authors has now responded to that post, and I to him. The more dialogue I get here the better the method looks, but I don't think the principals with which they tested it for this paper really gave it a good run.

Matthew Gabriele said...

Indeed, Jonathan. I'm well chuffed that my initial post has sparked such a discussion. I mean, how often do historians and mathematicians really talk to each other?

But, that being said, I still think we (historians) and they (mathematicians) are talking right by each other. They seem to think that we're critiquing their method but rather, I think, we're more critiquing their presumptions -- as you say, one shouldn't assume that 2 people are related just because their associated by the same notary. That's almost certainly NOT true in the case of King Philip I, where I've done some work, and most likely not true in the data set that they're using.

What this really all means though is that we need to talk to each other more to hammer these things out.