Limitations

6.4.1 Theoretical Limitations

Framing is a broad concept, which needs precise definition before being used. As evident in

Cornelissen and Werner’s work, the concept of framing has been understood in many ways and

subsequently applied to very different purposes. Cornelissen and Werner (2014) conclude, based on

the criticism of Entman (1993), that this widespread application of the concept has led to what is

described as a ‘fractured paradigm’. They do, however, conclude that the paradigm remains

90

relevant, but the ample definition is needed before the concept is applied. Therefore, we have tried to be exact in the way we applied and chose our framing method for this project. Nonetheless, we recognise that there were different theoretical options available to us.

We could significantly have altered the focus of this project had we chosen a different level frame. In our opinion, there were two other broad options available to us - we could either have chosen a micro frame or we could have chosen another meso level frame. It would not have made sense for us to use a macro frame, as we have found no way to relate it to Bluefragments.

Nonetheless, had we have chosen a micro level frame, we could have focused more narrowly on how the single employee makes sense of his or her surroundings. Cornelissen and Werner (2014) divide micro frames into either an analysis of the frame of reference - meaning the cognitive structure itself - or an analysis of the frame’s effects. We would then have focused on an analysis of the frame itself, but the effect would be outside our problem area. Furthermore, we could have used March and Simon (1958) or Starbuck (1983) as both analyses organisational culture and how an individual person frames the organisational culture. We would then have to analyse the individual culture in much greater detail and the project would have focused much more on the flavours of the individual clients’ culture rather than their understanding and aspirations for the use of AI - in other words, we would have lacked a focus on technology.

A more viable option would have been another meso level frame. According to Cornelissen and Werner (2014), technological frames are meso level frames and we could have applied yet another technological frame, as outlined by Davidson (2006). We chose to use Orlikowski and Gash (1994) as their focus allows for a broad analysis of perception. However, we could have gone the opposite direction and chosen a technological frame, which was more specific. As per Davidson’s categories (2006) it would have been most reasonable for us to focus on the frames related to information technology’s (IT) features or attributes, and also IT’s incorporation into work practices.

Had we used Livari & Abrahamsson (2002,) we could have made an analysis, which touched upon both of Davison’s categories.

Livari and Abrahamsson’s theory analyses the usability of a technology in the organisation.

They use both aspects of user-centred design and organisational culture theories in how the technology is being perceived by the different sub-cultures in the organisation. Had we used this method, we would have needed wider access to the client organisations as we would need more than one perspective from said organisation. We could have chosen a specific focus on a single client company to contrast the different sub-cultures within that organisation with that of Bluefragments.

We would have gained more complex understanding of how AI could potentially be perceived

differently within the same organisation and whether how AI was designed was also how it was

91

actually being used in the client organisation. However, we would not have been able to analyse how Ready-Made-AI had been used differently with other client companies.

None of the clients we have interviewed had fully implemented the AI yet. Had we found such a company, a third theoretical option would have been available to us. We could have moved away from framing theory and focused on use instead. We could have analysed how the AI had been interpreted after its implementation and how the members of the client organisation had coped with this. For instance, Stein et al. (2015) could be used to analyse the organisational members’

vacillating strategies. Nonetheless, it is uncertain how we would have contrasted the vacillating strategies of the client organisations’ members with that of Bluefragments.

In principle, we could also have chosen another interpretation of AI other than

socio-materiality as Leonardi and Barley (2010) presents four different approaches. In our project, it would not have made sense to use a determined approach, as we wanted to analyse how the technology has been interpreted in the various client organisations. A determined approach would have led to an unhelpful discussion about whether Bluefragments or their client’s interpretation of the AI was ultimately the correct one. We wanted to analyse the differences between the interpretations, not determine the best or most correct use of the AI. Therefore, the only option left open to us, would be social-constructivism, which works perfectly well to explain how people make sense of

technology. Additionally, it explains, to some extent, how people interact with each other around or through a technology. However, social-constructivism does not quite explain how a man and a machine interacts with each other, as it is not a social interaction as such.

6.4.2 Practical Limitations

Since it is Bluefragments’ clients that we are analysing, it makes sense that Bluefragments have been the mediator between us and the respective clients. Bluefragments have put us in contact with the organisations who were interested in participating in the project. The immediate implication with this set-up, is the fact that Bluefragments have had the opportunity to choose the clients which were already positive about Bluefragments and who believed that the collaboration had been successful.

All the clients we have interviewed believed the collaboration with Bluefragments had been a success, as did Martinsen. Our result might therefore be skewed towards an overly positive

perception of Bluefragments and the project. The presumption of our research question is that the

frame has been successfully established, but had we specifically pursued client negative perception,

if such clients exist, we might have been able to analyse the frame construction process in greater

detail.

92

As an example, in all three cases, Ready-Made-AI has been a successful concept, despite it not being clear to the client organisations. However, there could have been clients where Ready-Made-AI had been unclear and therefore resulted in an unsuccessful collaboration between Bluefragments and that client organisation.

We have chosen focused on three clients in our project as we are convinced that the perspectives of three client organisations is a minimum requirement for our comparative analysis.

We are of the opinion that more client organisations would have given a more varied analysis and result. However, we believe that we have a gained a critical mass of information in the interviews conducted for the analysis, which has given us a rare and specific insight into the collaboration between Bluefragments and their clients.

We might also have been able to improve the result of the analysis, if we have had wider access to the client organisations. We only had access to a single person in each client organisation.

Two or more interviews from members of the clients could have provided us with a more complex

picture of each organisation. Nonetheless, there would have been the added risk that the increased

complexity could potentially have led to a loss of direction for each organisational frame. In fact, we

might have had too much data to gain a clear picture of the client organisation’s frame.

93