4.1 Philosophy of Science, Research Design and Delimitation of Research Question
Before designing research, it is important to consider the epistemological and ontological approach to knowledge of a researcher. The epistemology relates to what constitutes acceptable knowledge, and I have adopted an interpretivist approach (Saunders, Lewis & Thornhill, 2007). This means that knowledge is based on interpretations of other’s actions and behaviours, because this will influence how we ourselves attribute meaning and carry out actions. Further, I followed a subjectivist ontological approach, which builds on the belief that “social phenomena are created from the perceptions and consequent actions of those social actors concerned with their existence”
(Saunders et al., 2007: 108). In line with this, I adopted a social constructivist approach to the research design, based on the assumption that reality is socially constructed and that, in order to understand subjects realities, I need to understand their subjective reality. This led me to design the research within the interpretive paradigm, allowing me to uncover irrationalities amongst actors, as well as rationalities, with a goal of understanding rather than changing the social order in which they operate (ibid).
The purpose of the study was to find out what was occurring with regards to digital agriculture in Brazil during the fall of 2019, whilst I was living in São Paulo. This was done by asking a series of questions to assess digital agriculture in Brazil, making it an exploratory study (Saunders et al., 2007). In order to do so, an abductive approach was adopted, using multiple embedded case studies to collect data for a cross-sectional contribution to existing theory (Dubois & Gadde, 2002; Saunders et al., 2007; Yin, 2014). The abductive approach allowed me to go back and forth between theory and findings, and incorporate new findings into an emerging and iterative framework. It was important for me to be able to be flexible during the research, as I was myself experiencing and learning new things about Brazilian culture, the importance of agriculture, and the social inequalities in the country.
4.2 Data Sampling and Collection
Seeing as the goal of the study was to explore the current situation in Brazil with no requirement for generalisability, qualitative data collection methods were chosen. The data was originally collected for the CEMS Research Project, as outlined in the Data Process (Appendix A). Sampling was done on the basis of both convenience, selection, and snowballing. My experience in a native community was a part of an organised seminar, which conveniently provided me with subjects from family-run farms. Following this seminar, an initial conversation with a professor led to the interview with the technology start-up representative, in accordance with a snowballing sampling method (Saunders et al., 2007). These conversations also made it clear that more than one type of actor needed to be considered, so I advertised the need for interview subjects amongst my network, and from there selected the medium-sized farm representative. Based on preliminary findings from the seminar and initial conversations, I created a topic guide for the two interviews which covered topics such as digital technology, environment, and the government, and also left room for emergent themes to be discussed (Appendix F). This topic guide was aimed at uncovering specific perceptions of the Brazilian context, as well as attitudes towards digitisation of agriculture.
4.2.1 Primary Data Collection: Interviews
During my time in Brazil I was able to attend talks and conferences by professors, journalists, social activists, and lawyers which provided contextual reference for the contents of my study. Interviews were chosen as a data collection method, because they provide insight into the discourse, physical movement, and topics of importance for interviewees (Easterby-Smith, et al., 2018). Semi-structured interviews were deemed the most appropriate, because of the flexibility they provide in discussion of topics, allowing me to gain insight into the the subjects’ attributed meanings to words and phrases, their ways of reasoning, and their perspectives of topics (Kvale
& Brinkmann, 2015). By conducting two semi-structured interviews with different subjects, I was also able to address the potential ambiguities between the responses (Saunders et al., 2007). During the interviews I was aware of my role as a researcher, making sure to ask open questions that related to the topic guide in order to avoid the influence of my subjectivity (Toma, 2000).
4.2.1.1 Group Interviews
In August 2019 I attended a seminar in the Atlantic Forest in the Southern part of the state of São Paulo, visiting a native community of Quilombo. During the two days my 18 fellow students and I spent in the village, I participated in several discussion sessions with three key persons from the community: the currently elected male leader, a political man in the community, and a woman growing crops for herself and the community (see Appendix D for pictures). The discussion sessions were structured around specific topics, and the group was able to ask questions for elaboration or other topics. The discussions were led by three Brazilian facilitators with backgrounds in business, anthropology, and sociology, who also translated the responses. The discussions were relatively unstructured, which allowed for the emergence of themes and topics that were relevant to the students and the interviewees. I made thorough notes during the discussion sessions, digitalised them upon return to São Paulo, and later analysed them for key topics and themes.
4.2.1.2 Individual Interviews
The first semi-structured interview was conducted with a representative, Vitor, from a medium-sized farm, located in the state of Bahia, in the North-East of Brazil. The interview was conducted via Skype with video at first, but due to technical difficulties and lack of internet connection, it was moved to Whatsapp and completed as a telephone conversation. Whilst asking questions from the topic guide, the interviewee also spoke about other actors in the field, providing insight into the context of their operations and field. The interview was conducted in English, recorded and transcribed.
The second semi-structured interview was conducted with the male founder of a Brazilian technology start-up, Luciano. The interview was conducted via Zoom, a video conferencing tool. Due to extreme rains that day, the connection was broken twice, but the whole interview was completed with video and audio. Like the first interview, it was guided by the topic guide, and emergent topics included the international environment of technology
development and academic theories on technological innovations. The interview was conducted in English, recorded, and transcribed.
Figure 2: Data Structure Digital Divide 4.2.2 Secondary Data Collection
Secondary sources of data are useful for triangulation, a method used to find support, or otherwise, for the existing line of inquiry (Barratt, Choi & Li, 2011). In this study, secondary data on external field actors was collected based on the emergent themes in the interviews. By using ‘active data’ (Dubois & Gadde, 2002), I was able to use the secondary data collected to form a bigger picture of the field than anticipated in the interviews. The sources of secondary data included the organisations’ websites and publicly available presentations. Almost all sources found were written in Brazilian Portuguese, and the technical phrases and statements that I did not understand were
translated using DeepL, an online artificial intelligence translation tool. Reading secondary data in the native language of the organisation provided useful insights into the phrasing and framing of the actor, and I was able to use my own knowledge of the language and the context to correct phrases where DeepL fell short.
Figure 3: Data Structure Framing 4.3 Data Analysis
Having adopted a social constructivist approach, discourse analysis was chosen as an appropriate method of analysis for the interviews. Discourse analysis allowed me to code the meanings of words and phrases that the respondents used, and compare them against one another. The initial data from the group interview led me to theory on framing, based on the fact that technology was seen as a negative development. Building on this theory, I conducted the two other interviews, which then led me to the literature on the digital divide (see Data Process Appendix A). The analysis was therefore done in retrospect, using both data and theory to code the findings.
emerged as significant in the data and in the literature. The data analysis was guided by a sensitised concept, namely digital agriculture, which acted as the lens through which I approached the topic (ibid). Due to the two different streams of literature applied in the analysis, there are two different data structures. The resulting data structures show how the data was clustered and combined to create aggregate dimensions (Figure 1 and 2). Within the digital divide literature, I found a lack of a coherent and overarching model for the analysis of actors’ engagement with technology. By combining literature and my findings, I therefore created a new conceptual model for the categorical positioning of an actor in the digital divide, which was the basis for my analysis in chapter 6.