Elite Network Shifts
Jacqueline Hicks, Vincent Traag, Ridho Reinanda, Gerry van Klinken, Fridus Steijlen, Ayu Purwarianti, Maarten de Rijke, Franciska de Jong, Andrea Scharnhorst, Peter Keppy
Elite Network Shifts – how did it go?
Elite Network Shifts is a ‘blue sky’ research project, marked by high ambition, a high level of interdisciplinary collaboration, and high uncertainty over outcomes. Its purpose is to use computational techniques to plot shifts in the networks among top Indonesian elites. The networks are read from a large digital news archive. Behind this specific goal lies an exciting bigger question – is it possible to automatically extract interesting sociological information from large digital textual archives? ‘Automatically’ means: without too much manual list-making or searching. ‘Interesting sociological information’ could include all kinds of trends involving people, places, events or ideas. ‘Textual archives’ could be newspapers or books over a long period (they do have to be digitised, which is an expensive chore).
At this stage of the project, it is appropriate to step back and begin to evaluate its overall achievements. Assessment should be as multidimensional as the ambition of the project.
• Scientific: were the hoped-for results achieved? The answer is, yes but only some. Of the three terms in the project’s title that indicated its ambition – Elite Network Shifts – only the first was successfully achieved. We did identify Elite individuals automatically from the newspaper texts. Computational researcher Ridho Reinanda was able to extract thousands of names correctly from the archive. Social scientist Jacky Hicks tested the ‘eliteness’ of these names and was first author of a paper on this in Comparative Sociology last year (see below). However, Networks calculated on the basis of the co-occurrence strategy (where individuals mentioned in the same sentence were assumed to have a ‘relation’) surprisingly turned out not to be sociologically meaningful. Mathematician Vincent Traag was lead author of several papers in international journals on this. The goal of identifying Shifting Networks (the historical aspect) was abandoned. We did not have the resources to preprocess large volumes of historical newspaper data in the Indonesian language. The goals remain worthwhile, but they turn out to be harder to achieve than we thought. In all core scientific areas, however, doors have opened to new ways of addressing these complexities. These can become pathways to future collaboration.
• Individual: did individual researchers publish? The answer is yes. All three core researchers published extensively throughout the project and grew professionally from doing so. Vincent Traag as a result won an exciting research position at the Centre for Science and Technology Studies, Leiden University, in November last year. Ridho Reinanda expects to graduate as PhD at the end of 2016. Jacky Hicks faced the biggest cross-disciplinary challenge, as she was operating well outside her professional comfort zone. She has become a sought-after presenter at international digital humanities conferences. Her ENS fellowship was extended so she could organise an international conference in May 2016 called Digital Disruption in Asia: Methods and Issues (http://www.kitlv.nl/digitalasia/).
• Collaborative: did the team work? Again the answer is definitely yes. The core researchers spent a lot of time communicating across their respective disciplinary barriers, and investing emotional energy to ensure the team stayed together. Considering that many cross-disciplinary projects fail precisely along this dimension, the team’s commitment was fundamental to any success the project achieved. One unexpected proof of this commitment is that, when the initial goals of the project seemed largely unachievable within the time frame, the researchers devised alternative goals. They drew up sub-projects: one on media bias and another on ways of automatically adding attributes to a database of elites. They kept working at these alternative goals even as time was running out. For example, the SIGIR paper by Reinanda et al (below) is about how to decide which attributes should be in such a database, based on what people are actually searching for on the web. And Traag’s PlosONE paper (below) is a study of elite relations in LittleSis, an American elite database that inspired our own (more advanced) database plan.
• Future: have new collaborative possibilities emerged? The answer is, probably yes. The project posed ambitious questions about automatically extracting historically and sociologically meaningful information from large, weakly structured, natural language textual corpora (in an Asian language!). This ambition does remain of burning interest to social scientists as well as to the computer scientists associated with this project (see the edition of Asian Journal of Social Science below). The current project has learned how difficult this is to do, but it has also learned ways of confronting those difficulties. We are discussing possibilities for future collaboration.
• Technical: has the project left a legacy of available tools and data? The answer is, hardly. The PhD sub-project in particular developed many tool prototypes, and proved their concept on the project data. However, to turn these into usable tools would have required a software support laboratory. Future collaboration can only work if such a lab is available for routine support. Its tasks can range from scraping, through preprocessing, to tool-completion, visualisation, presentation, and archiving. Copyright restrictions on obtaining and sharing newspaper data is another constraint on research sustainability. Finally, Asian Studies (the domain of this project’s social scientists) still faces huge challenges of basic digitisation, with few texts digitally available at this moment. It is worth investing in this.
ENS key publications
• Hicks, Jacqueline, Vincent Traag, and Ridho Reinanda. 2015. Old questions, new techniques: a research note on the computational identification of political elites. Comparative Sociology 14:386-401.
• Traag, V. A. 2016. Complex contagion of campaign donations. arXiv:1601.07679v1 [physics.soc-ph]. March 2016: accepted for publication in PlosONE pending minor revisions.
• Reinanda, Ridho, Edgar Meij, and Maarten de Rijke. 2015. Mining, ranking and recommending entity aspects. Paper read at SIGIR 2015: 38th international ACM SIGIR conference on Research and development in information retrieval.
• Klinken, Gerry van. 2015. Introduction: Digital methods in Asian Studies. Asian Journal of Social Science 43:539–44. (this special edition of AJSS also had papers by Hicks, Traag, and Reinanda).