academic history networks

William Godwin was a British political philosopher, writer and journalist. For 49 years in the 18th and 19th centuries, he wrote down his every social interaction. Over this period, he recorded 64381 instances of co-presence with other people, of which 49876 are with people who have been identified. Such a complete set of records is obviously of significant historical, political, and sociological significance.

But these data are difficult to analyse, both technically and theoretically. The graph below shows those connections which happen more than 40 times between Godwin and his acquaintances (here is an SVG version, where the nodes are hyperlinked to their biographies.)

Taking a subset of the nodes like this makes the data more manageable, but makes historical interpretation more problematic. We can't be sure to what extent a quantitative analysis takes into account the intention of the person making the record. In addition to all the other issues with self-reported ego-net data, we need to account for the varying reasons that interactions are recorded (for example, contacts with most — but not all — servants are not recorded).

I am currently working with Mark Philp to analyse the diaries in the context of what is known about Godwin from other sources. In particular, we are investigating gender-based patterns of interaction. There are surprisingly few sources about the general etiquette of calling on others in this time period, but then again, Godwin might be unique enough in his social beliefs that his social life can be analysed as sui generis.

However, more context-independent research can also be done, which treats the data as a generic egonet, ignoring any added historical information. While the initial coding of the diary viewed it as a literary work, the systematic transcription into TEI XML makes it possible to programatically analyse it. My current investigations focus on testing the applicability of Dunbar's number, and the use of directed non-uniform multihypergraphs. The most problematic part of the study so far is still working out which questions to ask...