ºù«Ӱҵ

The focus of our work is on how empathy and trust play out in social interactions around emotional distress in online spaces and, working jointly, we have come up with a number of potential ways to explore this. In this article, we share some of the challenges and opportunities of such collaborative computational social science from our perspective – having arrived from disciplinary backgrounds in the sociology of emotions and personal relationships (Julie) and internet and social media studies (Frances).

Transdisciplinarity or interdisciplinarity is best understood in the doing, and there are no easy ways around the potential messiness of that process. Like the family in the children’s book Going on a Bear Hunt, who encounter a muddy bog and a variety of other obstacles, we soon realised: ‘we can’t go over it, we can’t go under it – oh no, we have to go through it’.

---

Practicalities vs. meaning

One of the things we have found difficult as sociological researchers is knowing what is possible in terms of data science. Even if we know what we want to find out, we may not know how to go about this from a practical, computational perspective. So our work with KCL has been highly iterative as we explored the possibilities and limitations of what could be done.

However this process of ‘feeling our way’ through the data has led to a number of potential innovations in the research. Like the process of doing interdisciplinary research, the tensions of doing sociological big data research have been productive of new ways of doing things.

One of our major research interests as sociologists researching empathy in relation to emotional distress is the idea of empathic response. We conceive of the social in terms of relationality and interaction, but social media data collected simply through the use of keyword search terms often appears flat and divorced of social context.

For this reason, we collaborated to design approaches and potentially innovative data collection tools that have placed the focus back on response and on interaction.

---

Dealing with definitions

Sometimes it becomes clear that particular concepts are understood in different ways across disciplines. Looking at ‘emotions’ and ‘relationships’ from a sociological perspective, for example, is different to doing so from a data science perspective. Sociological research has developed a nuanced understanding of emotions as fluid and relational and not always expressible (see e.g. Burkitt 2014; Brownlie 2014).

This understanding of emotions sits uneasily with attempts at quantification of emotional and relational content through tools such as sentiment or social network analysis. Beside the linguistic limitations that inhibit these measurements (the difficulty of recognising sarcasm, for example), from a sociological perspective there are ontological and epistemological problems with such attempts to measure emotion and operationalise relationships.

For this reason, we have developed an iterative and complementary approach in which we move back and forth between the computational and qualitative approaches to the analysis of data, testing and re-testing our assumptions through this movement.

For example, computational approaches help us to narrow down or categorise our data according to a social practice, but we have more difficulty measuring content and interactions within that data. Through returning to qualitative approaches we can inform our understanding of how participants frame their participation.

Daunting though the process of ‘going through it’ has been, it has also been extremely rewarding not least because working with our computer science colleagues has been as revealing of ourselves, what we do and believe as sociologists, as it has about the nature of other disciplines and our relationship to them.

Authors: Frances Shaw and Julie Brownlie, 11 March 2016