Data Thematisation

In this section, we address the methodological approach to treat the collected data based on thematic analysis (Braun & Clarke, 2006). This section explains the approaches of processing the data that was gathered through the interviews, in order to indicate how we played an active role in

identifying the main themes of the data collected to conduct the study based on their interests in the research topic (ibid). It is vital to note that the identified themes are not explicit in the data but reside in our own minds framed by our hypothesis and approach to study. This means that the data corpus becomes subject to researcher judgement, interpretation and analysis (Ely et al., 1997).

In this thesis, the choice of strategy when analysing the data is a latent thematic analysis (TA). This means that the collected data will be theorised through examination of patterns appearing

throughout the various data items obtained in the shape of interviews of actors in the organisation.

Theorising the data means that the coding goes beyond the explicit semantic content of the data, but rather seeks to discover implicit themes recurring as patterns throughout the data set through our interpretation (Braun & Clarke, 2006). Our interpretation of the data is dependent on the theory that has created the foundation of knowledge for this topic along with preexisting knowledge about the particular organisation from previous collaboration. The prior collaboration project established some preexisting ideas and assumptions shaping the study. Therefore, it should be noted that the bias established precedingly to the initiation of the research undeniably influenced the conduct of this study. As previously mentioned, the aim of the first collaboration was to consider the influence from the introduction of Artificial Intelligence to the architectural profession and how this will change the practice of the architect. When doing this study, it became apparent to us that the case organisation works very actively with research, consequently forming the motivation for this thesis study.

The latent TA as a method to process the collected data brings forth a high level of flexibility and fluidity. This has its advantages as well and some disadvantages for the analysts. By disconnecting ourselves from the semantic content, where word count can be used to identify the patterns

(Bryman, 2012), the required interpretations of which latent themes are dominant is already initiating the analysis of the data . Listening through the entire data set of audio files of interviews several times is a crucial element of discovering and defining the recurring themes (Braun &

Clarke, 2006). The process of determining the required amount of recurrences and defining the pattern needed throughout the data set for these specific pieces of data to become a theme is also an element interpretation and selection. This was managed by transcribing all of the interviews and going through each individual interview to familiarise ourselves with the data corpus (see appendices). By doing so it enabled us to generate the initial codes. The initial codes were

discovered by marking out pieces of the data to then summarise the essence of the quotes through our own interpretations. These condensed interpretations of the data were then used as the initial codes. Subsequently to developing all of the initial codes, we started searching for themes, related to the overall research topic, into which the codes could fit. This required a collation of codes where

patterns were apparent. Collating of the codes led to comprehensive discussions about their relation which was an important part of defining our object of analysis for this thesis.

Thus, both the focus of the research and the object of analysis drifted along with the interpretation of the gathered data as the codes from the interviews enlightened new perceptions and

understandings of the topic (Braun & Clarke, 2006). As the development of the themes progressed there were several iterations required to see if the themes discovered were actually coherently connected and apparent throughout the entire data corpus. Through these iterations, the themes were subject to change and modification, for them to be more pertinent and inclusive.

In order to assess the coding and thematisation in a comprehensible way, an excel spreadsheet was created where the themes were listed horizontally and the codes inserted to the related theme (Appendix H). The codes were then colour categorised based on the interviewees to distinguish between the participants’ statements and the linked codes. This enabled a quantitative guideline by showing which themes were appearing more often than others and the prevalence of the theme across the data set. As Bryman (2012) argues, the identification of dominant themes can be done in a measurable way through quantification of the qualitative data. He proposes that frequency of phrases and repeated words may identify the most apparent topics and themes of the data. However, due to the interpretive nature of TA the proportion of occurrences were not considered in the

process of determining what should be a theme but rather if a theme contains vital information for the overall context. Hence, the judgement of the researcher is vital in the selection of themes (Braun

& Clarke, 2006).

By letting the codes generate the themes, we have worked with TA in a bottom-up way, making this an inductive approach to the TA (Braun & Clarke, 2006). However, with this being said, the

previous collaboration with the organisation had already created some preconceptions originating and the idea that innovation of architecture is influenced by a trichotomy of elements consisting of business, practice and knowledge. Moreover, the use of ANT as theory and method encouraged us to look into the actors, process and controversies in the network. By doing so, the themes that were defined should correspond to the purpose of the ANT which focus on the controversies. In this way, our choice of theory directly impacted our thematisation. ANT helped shape the development of the project and the focus of the analysis. Nevertheless, as Bryman (2012) argues, a tendency when

working with inductive reasoning is that deduction may influence the process. Researchers may seek to confirm or disconfirm throughout the process of data collection, which was also the case in this study.

The time pressure pushed us to initiate the data analysis before having acquired the full data corpus.

In light of the themes discovered in the TA and the relation between them influenced and modified the interview guide. Indeed, an element of deductive reasoning took place as we sought to confirm the initial findings from the TA of the previous interviews.

By mapping out the discovered themes, it enabled us to unfold the relation to one another. This led us towards rethinking how the analysis of this thesis should be carried out and structured. This visualisation, in the shape of a mindmap (Appendix I), made us decide that unfolding knowledge transfer as a narrative, explaining and analysing it in the scope of Actor-Network Theory would be the most interesting approach to the analysis. When drawing out the relations between the themes in the scope of our research question, it became evident that certain themes were dominant in this relation.

To conclude on the above, assessing the data has been an iterative process, where themes have firstly been identified, defined and compared with patterns across the data set. Thereafter, the themes discovered were theorised and contextualised with gathered literature before commencing the analysis. While writing the analysis, the subject of analysis also took shape as assumptions were formed by looking at the processes in detail.