Despite post-ANT and new materialism’s strong focus on the impact of research meth-ods (Savage 2013), there is surprisingly little literature that directly addresses how to carry out data collection and analysis using this approach. Typically post-ANT accounts combine a wide range of methods such as ethnographic observations, interviews and doc-ument analysis that are interspersed with high-level theoretical argdoc-uments. Yet it is often unclear why particular methods were chosen and how this contributed to the overall the-oretical argument. Papers within this approach that directly address methods (Nimmo 2011, Marrero-Guillamón 2015, Fox & Alldred 2015a,b) suggest that classic social science methods are not stand-alone instruments but need to be adapted to the epistemology and ontology of post-ANT and new materialism. Fox and Alldred suggest that methods need to be seen as part of a ‘research assemblage’ that combines “the researcher and the researched event, plus the many other relations involved in social inquiry such as the tools, technologies and theories of scientific research” (Fox & Alldred 2015a, section.3.1). In this argument, good research involves gathering a sensitive combination of elements includ-ing the researcher, their theories and methodological tools in order to allow differences to be detected. Fox and Alldred describe the research assemblage as a ‘data collection ma-chine’ that needs to be sensitive enough to detect variations within the phenomena under study as well as being able to affect them. My study adopts this approach of treating in-dividual research methods as part of the whole device study assemblage. This means the
methods of ethnographic observation, interviews, surveys, document analysis and design prototyping are used in conjunction with my own background knowledge and context to tell the narratives of the devices. I also adopt an approach to ethnography from Marrero-Guillamón (2015) who describes narrating a “multiplicity of possible points of view; conse-quently, discovering and adopting such perspectives” (p.14). He provides an example of combining interviews with different people to construct them into a singular spokesper-son that can speak on behalf of an object and narrate its ontology. Figure 3.2 shows how the four research questions of the study are addressed by the research methods.
1 What is being sensed & what is doing the sensing?
2 How do power and politics take place within participatory sensing?
3 What can ontological design offer participatory sensing?
Horizontal Analysis Analysis across all the devices to address the
4 How does participatory sensing construct sensation and the environment?
Figure 3.2: Methodology diagram showing how the four research questions (green) are addressed using a variety of data collection methods (blue) that are applied to the different device studies (shades of orange) and Heathrow prototypes (purple). The white boxes describe each method.
Ethnographic observation
The ethnographic observations were carried out during formal and informal meetings
and events across the three-year duration of the EveryAware research project and sub-sequently during the design prototyping in Heathrow. The research question of ‘what is being sensed and what is doing the sensing?’ is a material-semiotic question that requires observation of material and human interactions. So the role of these ethnographic obser-vations is to record the richness of micro narratives of the devices, whilst paying attention to my own and other people’s interpretations and feelings as events are unfolding. This approach involves an awareness of auto-ethnography, but the aim is not personal experi-ence (Ellis et al. 2011) or notions of subjectivity (Butler 2009); rather the intimate dynam-ics with the devices. Thus a lot of focus is on the way the devices are being configured and used and how this is reflected in the claims made about the devices. This approach is inspired by Lucy Suchman’s studies where she suggests paying attention to “how they talk among themselves and with relevant others, how they translate their own embodied courses of action into written accounts and other materializations, and how they assess the meaning and adequacy of materials created by others” (Suchman 2000, p.312). These observations were then noted down as short first person research diary entries and stored as a series of individual text files on a laptop. The research diary entries are accompanied by multimedia material such as photos of devices being used and screenshots of visuali-sations. Some of the consortium events and internal meetings were also audio-recorded with the consent of the participants. Additionally material such as email communications were used to trace temporal patterns in discussions. After preliminary testing of this ap-proach in the WideNoise study, this method of organising observations was standardised across the other device studies. Here is an illustrative entry from the research diary:
“Whenever I talk to [name] about helping calibrate the device, he always sends me these long complicated formulas that he doesn’t seem to fully understand.
When we tried to implement the formula it all seems to rely on a magic voltage level that can’t quite be explained”.
An analysis of this extract highlights the way calibration acts to provide legitimacy even though the material basis is often blackboxed. Due the long duration of the studies, it was possible to track changes in these kinds of observations, such as the topic of calibration disappearing over time. These temporal patterns were analytically important for indicat-ing that an ontological shift had taken place. These observations were then discussed with three other researchers from the EU funded project to triangulate these interpreta-tions. Such changes were analysed using a relational mapping approach and visualised
using the OmniGraffle software to identify patterns and connections between entities.
The write up of the devices studies uses a ontological mapping method to illustrate the ontological relationship of actors. This technique was adapted from Fox (2017), who pro-vides an example of mapping the connections of actors around an insulin pump, “blood sugar – insulin – diabetes – pump – user – clinical specialist” (p.141). This mapping method was used to analyse the ethnographic observations and clarify the structure of ontological changes across the device studies.
Material analysis
This method is focused on the material properties and affordances of the sensing devices.
It involved detailed analysis of the source code of the WideNoise app in order to under-stand how the calibration function translated measured voltages from the microphone into decibel numbers. It also involved experimentation with the AirProbe device such as carrying the sensor box for a week to see how it affected my way of being in the city.
This kind of material analysis was important for being able to understand how sensitive the devices were and how they would act within a real-life usage. This proved critical in order to have informed and detailed discussions with participants about their usage of the devices.
Interviews
The study makes use of two kinds of interviews. The first were structured interviews with participants within the remit of the EveryAware research project. For the AirProbe deploy-ment in London all the participant groups were individually interviewed using structured approach with a fellow EveryAware researcher. These interviews were then transcribed and thematically analysed and triangulated. The focus of the interviews was on the goals of the EveryAware study such as motivation for participating and impacts of the activity.
Yet these interviews were also used for my own ethnographic study and the respondents were informed prior to the interviews, that the data would be used for both of these pur-poses.
The second interviews carried out as part of my ethnographic study were much less struc-tured and more responsive in terms of content and timing of interviews. As an example, I carried out an interview with a researcher of the Green team in a cafe after the final evaluation of the project, while we were waiting to catch a taxi to the airport. This respon-siveness was crucial for being able to capture an immediate and emotional reflection on
the AirProbe device that would have been lost otherwise. During these responsive in-terviews many different topics would be discussed without set questions. I saw my role as interviewer to reflexively introduce new concepts to the discussion and confront the respondent with contradictions. Whenever possible I would have the sensing objects physically present on the table while the interview was taking place, so that both inter-viewer and respondent would stay focused on the device and could handle and gesture at the object. This responsive and performative way of using interviews functioned within the broader approach of the device study where the researcher and context were a crit-ical part of creating a sensitive research assemblage. Fox and Alldred suggest that in a new materialist approach “interviews must be treated not as means to obtain subjec-tive representations of the world but as evidence of how respondents are situated within assemblages” (Fox & Alldred 2015b, p.409). Essentially, interviews are not about pinning-down people’s viewpoints but about narrating the broader practices that the respondent is involved in. Here is a short extract from one of the interviews:
Interviewer: “I was watching your face when it was flatlining at zero. [Inter-viewer laughter]. During the demo it was recording zero. It is just sad”. Re-spondent: “You can tell it to these guys. I was a bit surprised by the American guy, his background is in measurement”.
An analysis of the extract highlights the way the malfunction of the sensing device be-came an occasion for both humorous emotion and intensity that emphasises the ambi-guity of the data generated. Due to the quantity of interviews, around 20 where fully transcribed, while the majority were analysed by identifying key sections for closer anal-ysis. The coding of the interviews was carried out using a thematic clustering approach using the Scrivener software that allows textual chunks to be modularly re-ordered. I ex-perimented with a structured coding method using ATLAS.ti, but this turned out not to be suitable since the level of close textual analysis was not appropriate for the wider research assemblage. The looser thematic coding in Scrivener allowed me to identify broader pat-terns that were then discussed and triangulated with fellow researchers.
Document analysis
This method is used to analyse the broad range of documents such as websites, videos, funding proposals and mailing lists that surrounded the sensing devices. Nimmo (2011) suggests that “ANT offers a distinctive way of seeing texts which challenges the stan-dard ethnographic view of texts and fundamentally transforms the issue” (p.108). This
means documents are not static representations but inscriptions that can act as actors in reshaping practices. This study uses this approach for documents such as the Kick-starter campaign videos of the Air Quality Egg and Smart Citizen Kit, which were critical to the success of gathering financial backers. Analysing these documents as anticipatory inscriptions provided critical insights into the intentions of the devices (Law & Ruppert 2013) in relation to issues such as accuracy, participation and distribution of sensation.
The thematics of the documents were then analysed with a tight focus on the indicator of ‘what is being sensed and what is doing the sensing’. These observations were then triangulated with fellow researchers.
Surveys
The study makes use of pre and post participation surveys of AirProbe and WideNoise created for the EveryAware project. The surveys were developed in conjunction with a fellow EveryAware researcher and were intended to gauge the motivation and change in environmental understanding of the participants. This thesis does not make statistical claims based on these surveys; instead, they are used as components of the research assemblage and in conjunction with the ethnographic observation method.
Design prototyping
This method involves the iterative prototyping of experimental devices to act as alterna-tives to the existing sensing devices. This involved the development of custom electronics and software. This method was used to revisit the Heathrow site and design a series of custom prototypes to investigate the infrastructure of the local controversy. It involved staging and facilitating workshops where the devices were explored and designed via a participatory design process. The prototypes functioned as interventions and provoca-tions for local residents in a workshop with an aim of going beyond the existing framings of environmental pollutants. In this way, this method became a form of ontological de-sign where a participatory sensing device could create new material manifestations and realities of noise. The discussions between participants of the workshop were recorded as audio files for transcription, coding and analysis. The overall process of prototyping was recorded in the research diary.
Ethical Considerations
When I started this research process there was no practice of ethics approval within my re-search department at UCL. So I used my own procedure in line with the British Sociological
Association (BSA) ethical guidelines (British Sociological Association 2002) and with an awareness of ethnographic good practice (Murphy & Dingwall 2001). I respected the phys-ical, social and psychological wellbeing of the respondents in this study and minimised any possible harm due to this study. Most of the ethnographic fieldwork recounted in this study took place more than three years ago over the period 2011–2014. This time-delay between the fieldwork and publications lessens the ‘temporal positioning of risk’ (Murphy
& Dingwall 2001, p.340) for the respondents of the studies, since the sensing technologies discussed have now been replaced with newer technologies.
I have anonymised all the respondents in the study in order to ensure their privacy and confidentiality. Respondents are referenced only via their role e.g. ‘the resident’. The academic researchers on the EveryAware project are referenced as belonging to one of five different teams that are anonymised as Red, Blue, Green, Yellow and Orange teams.
Sometimes researchers are referenced by the disciplinary context of their institution when this is critical for the argument being made.
All the respondents provided informed consent after an explanation of the goals of the re-search study, the rere-search funder and how the rere-search will be disseminated. It was made clear that people could withdraw from the study whenever they wanted. Where respon-dents were involved in both the EU research as well as my study, it was clearly explained to respondents that the data would be used for both purposes. This process of consent was ongoing and responsive. There was some opportunity for the respondents to challenge and clarify my analysis of events. The final text discussing the two EveryAware devices was formally submitted to two researcher from the EveryAware team for comment and clarification.
During formal interviews, the consent procedure was audio-recorded at the beginning of the interview. In collective meetings that were audio-recorded, the respondents were notified before each meeting and the microphone placed in a prominently visible position for all to see. An example of having to adapt this approach was the request from one of the EveryAware researchers that certain meetings should not be audio-recorded, which was immediately adopted.
The study does not communicate any private or confidential information and no personal email conversations are used within this study. The vast majority of the data that is cited
in the AQE and SCK studies is publicly viewable on the project websites and open mailing lists.
The research data was treated as strictly confidential and handled in accordance with the provisions of the UK Data Protection Act 1998. All the data was stored on an encrypted laptop and all data backups themselves encrypted.