Data and Displacement Case Study

Data and Displacement: Promises and Challenges of Open Qualitative Data

Case study 2: The Data and Displacement Interview Archive

By Timothy Monteath

What data are we talking about?

The Data and Displacement Project data archive was created by an international research team based at the University of Warwick, the University of Ibadan (Nigeria), the University of Juba (South Sudan) and the International Organization for Migration, which is led by Professor Vicki Squire. This archive holds a qualitative dataset comprised of transcripts of interviews conducted with internally displaced persons (IDPs) residing in camp-like settings across North-eastern Nigeria and South Sudan. The data archive also includes interviews with stakeholders from each region and experts in data management from the humanitarian sector, though these interviews are restricted due to confidentiality. All of the interviews shared with the UK Data Service are anonymised, with identifying details having been redacted by the research team. This dataset provides a foundation for interpretive research, enabling the exploration of the opportunities and ethical dilemmas inherent in open qualitative data within humanitarian contexts.

An important aspect of the Data and Displacement project was the visualisation of the humanitarian 'data journey' (Bates et al, 2016). To understand the benefits and challenges of open qualititative data, we can consider the data journeys of the transcripts created by the research. These were coded by the Data and DisplacementLink opens in a new window research team at the University of Warwick, following interviews conducted by the broader team between October 2020 and September 2022. The project assessed data-driven humanitarianism from the perspective of internally displaced persons (IDPs) in two contexts that are characterised by conflict and high levels of displacement: north-eastern Nigeria and South Sudan. It was jointly funded by the UK Arts and Humanities Research Council (AHRC) and the UK Foreign, Commonwealth and Development Office (formerly UK Department for International Development – DFID), under the Collaborative Humanitarian Protection Programme.

The project examined the production and use of large-scale data targeting processes in each context, focusing on the operational and ethical challenges that arise from the collection and use of such data. The team used mixed methods, combining a range of data analysis techniques with qualitative analysis of in-depth interviews.It sought to involve local stakeholders in the assessment of data-driven processes of targeted assistance, in particular IDPs themselves.

The collection of transcripts of this study was subsequently uploaded to the UK Data Service (UKDS), and is available for access by interested researchers registered with UKDS. All requests are subject to the permission of the data owner or his/her nominee.

How does this data enable open interpretative research?

As a data archive of interview transcripts, the Data and Displacement project raises wider questions of how qualitative research can be made 'open' for further interpretation. Compared with the advances of open social science in quantitative disciplines, interpretive work is still in the methodological 'foothills' of opening up research. As became clear during a data workshop that we hosted at the University of Warwick in July 2025Link opens in a new window, "open interpretative research" can mean many different things. For some researchers, the label does not just refer to the easy availability of qualitative data for secondary data analysis, but also suggests that the findings or conclusions of this type of research are not closed or fixed and can be revised by posing new research questions.

This interpretative perspective on open qualitative inquiry emphasises that social research cannot easily be separated from the perspective of the researcher, their discipline and their culturally specific concerns. Within such an approach, making the dataset open must be first and foremost understood as a way of extending the process of data interpretation beyond the research project that generated this data, and in doing so the original researchers cede control over this process. Importantly, the Data and Displacement project itself questions the current urge for more and more data for the targeting of humanitarian assistance to IDP. In evaluating the Data and Displacement project in this case study, we ask how the meticulous and innovative research of this project fits within Sabrina Leonelli’s (2022) concept of open data that seeks to further 'judicious connection'.

How can "judicious connections" be created through open interpretative data?

The wish to 'open' up a qualitative data archive does not automatically ensure that it will be useful as a source for future enquiry. The work of Sabrina Leonelli (2022) draw's attention to exactly this problem by asking what data is being opened up for? She makes the distinction between, on the one hand, an infrastructural approach in which technically sharing data openly is sufficient for open (social) science and, on the other hand, open data practices that seek to foster a 'judicious connection', in which the relational, situational and equitable (re)uses of data are prioritised. One of the main challenges for qualitative research in this regard is the loss of the context when the original research team is removed from 'their' dataset. This is especially challenging in interview-based research, where the importance of the tacit, situational and relational knowledge of the researcher to developing capacities for interpretation is widely recognised. When publishing a set of interview transcripts there is an inevitable loss of this embodied knowledge of the qualitative researcher, creating a 'detached' secondary source.

The research outputs created by the original research team by analysing the data themselves, of course, involve the placement of the research within a wider set of relations too – in conversation with specific scholars, the wider discipline, interdisciplinary concerns and much more, as the published research results travel outside of the concerns of the academy. Consequently, one of the themes discussed during our data workshop was what can be done to help contextual information ‘travel’ with the data on its journey. The publication of a qualitative data set itself could benefit from publishing a data paper to accompany the interview transcripts. Such a data paper could encompass: explaining the metadata, reflections on research positionality, the decisions and rationale behind interview directions, and the information given to the participants about the possible re-use of these transcripts, or discussion and themes that, while not present in the transcripts themselves, may help contextualise their reuse.

To prioritise the relational, situational and most importantly, equitable reuse of qualitative data also requires that we should try to anticipate the ways in which data could potentially be re-used. Given the hurdles that interviewees have had to overcome to participate in the research, care needs to be taken in how their stories are disseminated. Indeed, the data access rights for the Data and Displacement Project through the UK Data Service (UKDS) and the filtering of all requests through the data owner or his/her nominee is in place for this project and ensures exactly this. To further explore this challenge, we can consider how data could travel in a way that is outside of the scope of the original research, yet may nonetheless be of use in furthering interpretive enquiry in the mode of "judicious connection."

A different way of opening up Data and Displacement data: considerations

There is no doubt that the stories of displaced people are extremely important and deserve to be made public and to be known in different parts of the world. Yet, one of the challenges of opening up any dataset is that it is impossible to anticipate how it may be of use to future researchers. While it may be guessed that common academic forms of coding and recoding interview transcripts could be used by future researchers, other unexpected methodological innovations or research questions may enable new ways of working with the data.

We experimented with developing such a previously unexpected transformation by using generative AI to create a comic strip from the transcripts (see below). Rather than seeking to ask what can we learn from the data itself, we sought to ask what can we learn about a tool for data re-use through using it. With the help of the designer Ginevra Terenghi, we tested using ChatGPT to convert the transcripts into dialogue tabs that reflected them as interviews and research dialogues (see the Figure below).

Using the data in this way quickly revealed the limits of ChatGPT and Large Language Models (LLMs) more generally as tools for the 'creation' of new images. By inferring the context of the interviews, such a transformation reveals what it is that AI models and ChatGPT, in particular, are able to ‘interpret’. More accurately, perhaps, this could be described as ‘hallucinatory’ re-inventions, which display all of the flaws we have quickly come to expect from AI, from extra hands in AI generated paintings to highly biased representations of settings that are highly colonial and paternalistic. The prototype comic strip we created is thus first and foremost a demonstration of the limitations and potential harms of using AI models to transform and interpret qualitative data.

Such a transformation also stretches what it can mean to do interpretive enquiry with qualitative data. The data has travelled far from its original purpose, yet may be of use in speaking back to the original aims of the research. As the extractivist ‘data revolution’ in humanitarianism and the paternalistic dynamics that humanitarian data furthers (Squire and Alozie 2023) inevitably turns towards the incipient 'AI revolution', uncovering and testing the potentially extreme biases baked into blackbox LLM models becomes ever more important. Is such a critical demonstration of the risks of AI-based data reuse, however, a way to demonstrate, through a counter-factual, what a ‘judicious connection’ that favours the relational, situational and equitable reuses of data requires? How can such connections be more positively anticipated and fostered as data travels?

Re-orienting data for future interpretative enquiry?

Based on the above, we offer the following recommendations to researchers with an interest in making qualitative data such as transcripts of interviews, conversations or focus groups available for open interpretative research. Our recommendations are drawn broadly from the curating data cultures project and may be relevant to a variety of open qualitative datasets:

It is valuable to include as much information as possible about the context of data creation by publishing a data information sheet. So that, together with the transcripts, more details of where they occurred and under what arrangements can be made available.
Attach other supplementary information that may be of interest and use to future researchers, such as: the participant information leaflets, blank consent forms (already included in the UK data service repository), or descriptions of general conversations of context prior to the start of the interviews.
Consider the positionality of the researcher as part of the 'metadata' in opening up the datasets for interpretation and the implications this has for the people who give their stories and for future researchers.

Our second set of recommendations focuses more specifically on thinking through who has access to the data, how data can travel within interpretative enquiry and how researchers can think through how their data could be reused.

As the data resides with the UKDS and access requests are subject to the permission of the data owner or his/her nominee, what does this mean for future research? In such a context using a more restrictive approach to 'opening up' the data may, precisely, contribute towards fostering a 'judicious' open social science that it is better able signal how and why this data should be reused. Defining what such conditions are could help to ensure that future reuse aligns with the original research question – indeed, this could be a specific condition for access.
Similarly, are there methodological approaches which are favoured or those which should be ruled out as inappropriate? Defining a ‘code of conduct’ which specifies how and when access could/should be granted further opens up the process of sharing this data and introduces additional transparency into the process. Furthermore, it may make it easier for the data owner to make decisions on granting/denying access to the dataset, for a nominated data controller who may not have the depth of understanding of the dataset as the original researcher, and enable other scholars to think creatively about how they could use the dataset for future research.

Feedback from Vicki Squire & team

This is a fascinating reflection and example how the Data and Displacement interviews might be reused using AI, which raises a series of important questions that have been at the back of my mind when uploading project data onto the UKDS.

Before addressing the specific suggestions, I think it is important to note that not all our interviews have been shared for re-use. Our interviews with humanitarians working in international agencies and INGOs have been excluded due to their having undertaken by members of our team who are 'insiders'. This meant that the level of disclosure was specific to this relationship, which would have been compromised by a request regarding the reuse of data.

There are a few issues of concern worth flagging here. Firstly, it is troubling that we did not upload interviews undertaken with our largely white research participants from the 'Global North' but that we did upload those undertaken with sub-Saharan African communities from the 'Global South'. This hints at the difficulties of moving beyond extractive research processes with disadvantaged and racialised communities, even whilst the team was consciously working to challenge these dynamics. A situation in which social science researchers are increasingly under pressure to justify 'exceptions' to the norm of data sharing raises concerns about the broader structures of power and inequality within which such a norm is embedded.

Moreover, it is critical to recognise that social science research often relies precisely on relations of trust that are forged over time, and requesting the future use of data simply forms a barrier to such relations. This is the case for a range of research participants. On other projects, I have built up significant trust with people in situations of precarity and displacement, which renders the sharing of data for reuse problematic ethically as well as in terms of on-going relations of trust with research participants. Again, this is to raise caution about the shift toward the norm of data sharing, particularly when it comes to qualitative research that has not be generated with this as a key goal.

Beyond this hesitation about the increasing demands around data sharing and the risks of this in compromising the integrity of research relationships, the question of 'judicious connections' is a very interesting one. Our positionality as researchers and the context within which the research was undertaken is indeed an important aspect to forefront in this research, and is something that we have always tried to write into our own publications. So, it does seem rather perverse that the data is accessible without this contextual information. This is a key reason why I have always maintained the requirement for people to contact me as PI before reusing the data, because the expectation that the data can be used 'blind' is problematic in my opinion. I work on the assumption that a discussion about the further use of data, such as that in our project, is necessary to check that the purposes of reuse align with the project's original aims. Interestingly, I have never had to do this and the current dialogue is the first case when the reuse of our data has arisen for negotiation. There are, of course, risks here, because the negotiation process can be time consuming or could potentially also raise tensions and disagreements.

The comic strip is a great intervention and I thank the team for sharing this with us. We also created visualisations in our project based on our engagements with research participants, and interestingly we also had similar feedback questioning the assumptions that appeared to be 'baked into' some of the team’s initial visual representations. These were precisely a tool for further engagement with research participants of the type that the project sought to achieve. Perhaps this method of reusing the data can also therefore play an important role in extending reflections on 'judicious connections' in ways that that the original project team may not have considered. There are lots of factors that work against this sort of ongoing work, not least resource limitations, but finding creative ways of so doing seems to me to be an important and compelling endeavour. Thanks again for this opportunity!

Vicki Squire, PI Data and Displacement