The emerging field of digital humanities is the latest discipline to benefit from big data analysis and in an unusual arts-Stem collaboration, academics at UCD have used it to reveal new insights into the 2009 Ryan report on institutional child abuse.
The cross-disciplinary team behind Industrial Memories are Emilie Pine, associate professor of modern drama at UCD, Prof Mark Keane of UCD's insight centre for data analytics, and research fellow Susan Leavy who began work on the project in 2015. In Emilie Pine's view, the Ryan report is "probably the most important publication in the history of the State, yet we're not reading it. A lot of the material is witness testimony in the form of letters, diaries, memos and record keeping books. To me, it's the most important part of the report and I wanted to be able to read it and make it accessible to others. However, that's not so easy with a report that runs to 2,600 pages."
Digital humanities bring artificial intelligence and text analytics to bear on traditional arts and humanities scholarship. These techniques are already in use in business where big data applications enable companies gain insights into consumer behaviour patterns, for example. But now they are being applied to the humanities and what’s emerging is a new way of probing texts that uncovers things you cannot see or appreciate from a traditional surface read.
In the case of the Ryan report, the techniques enabled Pine to sift forensically through material and to make connections that had remained invisible during normal perusal of the text. For example, she was able to track and tag every interaction between all of those involved. Individually these interactions looked isolated and innocuous, but when they were all drawn together in one place it became clear that there had been deep and widespread awareness of the abuse.
Pine approached the Industrial Memories project from the perspective of witness studies; a field of scholarship that aims to ensure traumatic events in a country’s history, for example, are not forgotten and that the lessons learned are incorporated into society and academic research. However, she knew they could not rely on people to read the lengthy report. This is where Mark Keane and Susan Leavy fitted into the picture. Over a two-year period, they turned the report into an easily searchable database that shed new light on what had gone on.
‘Clear visualisations’
“With clear visualisations we show, for the first time, the active networks behind the industrial schools,” Pine says. “This gives us a picture of how abusers were transferred between schools. We can also see how people within the system communicated, including parents, the religious staff, and the Department of Education. These visualisations nail the lie that people – and the Government – did not know what was happening in these institutions.
“The close textual analysis also gave us new insights into the experience of abuse as co-ordinating the material uncovered another type of mistreatment which took the form of a dreaded anticipation of “waiting” to be beaten or assaulted,” Pine says. “We were also interested in getting to the heart of who knew about the abuse because one of the big issues for survivors is this general misapprehension that people outside these institutions didn’t know it was going on. To investigate this comprehensively we built a social network that logged every moment of communication between the key actors – residence managers, the Department of Education, parish priests, parents, local TDs and so on. It became very clear that people did know and the biggest node on the network was the Department of Education. I feel this was underrepresented when the Ryan report was publicly launched. The focus has been overwhelmingly on the religious orders when the responsibility is actually more widespread than that.”
Dynamic search
At a very practical level the project has created a dynamic search function that allows people to interrogate the report in specific detail for the first time. It has also created a “people directory” so that those involved can be identified by readers albeit it through pseudonyms.
“We don’t just take existing digital technologies as they stand. There is a lot of innovation and redesign involved in applying artificial intelligence within the context of humanities research,” explains Susan Leavy whose career began in the arts before she branched into technology and specialised in artificial intelligence. Prior to joining the Insight Centre Leavy spent 10 years on the technology side of investment banking. “The aims, purpose and scale of data are different to industry and require a different approach to the technology. The main challenge is using the tools of science while maintaining the critical and interpretative rigour of traditional humanities research.
“With something like the Ryan report, which is a long narrative with facts and figures buried in the text, the power of data analytics can be transformative. It provides new ways of finding information and makes texts much more accessible,” she says. “We used a combination of technologies including neural word embedding, machine learning and social network analysis to deconstruct the Ryan report and uncover patterns in the data.”
String searches
Prof Mark Keane is a psychologist by training whose interests include artificial intelligence, cognitive science and connecting computer science to humanities. “We took a set of PDFs of the Ryan report and imported them into the Django web framework, which meant we had all the chapters in a format that allowed us to do string searches. We were also able to pull out all of ‘the actors’ and create a category system that allowed us to navigate the report more easily,” he says.
“One of the most useful things we did was a transfer analysis that allowed us to see how priests were constantly moved on when there was a problem. We were able to create diagrams that showed this happened yet it was something you would never have realised from an ordinary reading of the report. The report was organised by institution so tracing the movement of one individual manually would have taken a very long time. Now you can do it in seconds. The real ‘overhead’ in a project like this is the database back end and the time spent designing the categories and relating them together in a database schema. To my mind, the really exciting thing about being able to analyse a very large body of text is its ability to show you things that are not otherwise apparent.”
The Industrial Memories project, which was funded by the Irish Research Council New Horizons, can be accessed at https://industrialmemories.ucd.ie
