3. Research Design and Methodology
3.9 Bias and Insider Role
Working as a manager of international activity at one of the studied academic units as my primary position and as a faculty member of the same academic unit on part-time
employment, I recognized that my insider role intersected with the purposes of the present study. Another fact that influenced the perspective of data analysis stemmed from the fact that the research and the interpretation of its results were done by only one person. Therefore, it was important to recognize the risks of bias and identify a solution for limiting them in the present study.
Bias is defined as a “systematic error introduced into sampling or testing by selecting or encouraging one outcome or answer over others” (Merriam-Webster, 2022). According to another definition, with respect to qualitative studies, bias is “any influence that distorts the results of a research study. Bias may derive either from a conscious or unconscious tendency on the behalf of the researcher to collect data or interpret them in such a way as to produce erroneous conclusions that favour their own beliefs or commitments” (Bloor & Wood, 2006, p.21). Bloor and Wood (2006) also write that bias could happen at any stage of the research process from research design to report writing.
The concept of bias originates from the quantitative research paradigm (Galdas, 2017). In quantitative research, the necessary actions to decrease the bias are clearly defined, i.e., they include statistical tests to measure the size of an error. However, such actions cannot be applied to qualitative studies. The qualitative research paradigm offers other methods to decrease bias, such as respondent validation, prolonged involvement into study, independent analysis of data by other researchers, triangulation of research methods and others (Smith &
Noble, 2014).
Below I describe what actions helped me reduce potential bias. First of all, in reference to the whole study process, since I was working in an academic unit and were not part of the central division of the university, I was able to achieve a certain degree of detachment from the studied institutional aspects between my researcher and professional roles. Since most of the interviews were held in other academic units or at university, this
also put me in a less-insider position. Additionally, during my PhD studies I went on
maternity leave; at this time, I was conducting the interviews with the study participants, and, in particular with the interviewees from my own research unit. I believe that being
temporarily away from the academic context allowed me to increase the distance between my role as a researcher and my colleagues as study participants. Furthermore, other specific actions that I undertook at various stages of my research in order to decrease the risks of bias, are indicated in Table 10.
Table 10
Risks of Bias and Responses to Them
Risk of bias Potential effect on data/findings
Actions to manage risk Design bias
happens when
researcher’s preferences or beliefs influence the choice of research methods rather than the optimal choice that works best to answer research questions.
Incomplete and incomprehensive response to the chosen research questions due to inappropriate research design
The research design was revised twice together with my research supervisor.
After the first revision, I changed the scope of the research due to the data access.
After the second, I added positioning theory as a theoretical lens since I had noticed that the response to my research questions would be incomplete without it.
Selection bias takes place when a researcher excludes a part of population from the scope of the
research
The selection of three academic units makes it difficult to
generalize the results at university level.
I included the institutional level into the research design, in addition to the academic units:
interviews, RPE and analysis of context were carried out at university level as well.
Additionally, prior to the selection of academic units, I collected data from two departments responsible for monitoring UrFU’s indicators of internationalization, and discussed the choice of academic units with their heads and my research supervisor.
Table 10 (continued)
The selection of study participants that are the most active players in
international processes makes it difficult to generalize the results at academic units’
level.
I realized from the very beginning of my study that the voices of this research belonged to the most active employees of the chosen academic units in terms of international activities. I made an effort to make this cohort of participants as diverse as it could be and invited the
representatives of the following groups to take part: domestic faculty, international faculty, leaders of academic departments/international study programs/international research units, heads of international relations, heads of academic units.
I also asked directors and insiders of two academic units, where I did not work, to help me form the list of employees to invite to the interviews in addition to the list of prospective study participants that I compiled on the basis of open sources.
The selection of interviewees in my own academic unit could be biased by my ideas on whom to invite, and some important voices could be missed.
After creating a preliminary list of study participants, I discussed it with the head of international affairs office and leaders of academic departments participating in the research in my own academic unit. It allowed me to add more people to the list whom I did not intend to invite at the beginning.
Data collection bias occurs when the researcher’s personal point of view on the studied questions influences the way information is collected.
The risk of missing data existed for document analysis and close-end RPE.
In order to reduce the influence of my own beliefs, I developed predetermined themes and rubrics for document analysis and tried to address a unified list of documents for each academic unit. I also used a close-end
questionnaire for the analysis of rationales that made responses more unified.
At the same time, close-end and predetermined collection of data generated a high risk of missing the ideas that were out of the scope of these two methods but were important for my case study. That is why I held open-ended interviews with the study participants in a semi-structured format.
The risk of biased data collection via
interviews was high as my insider position could influence the interview process through the choice of questions, wording and the whole interview process.
I used a preliminary interview guide which was unified for each group of study participants. It allowed me to control the focus of interviews, as well as to keep the range of questions and their wordings available to me during the interviews.
Table 10 (continued)
My insider role might also influence the answers of the interview participants in various ways: it could affect the readiness of my colleagues to participate in the research, as well as the way they provided responses during the interviews.
I arranged all the interviews to take place at an agreed time and place. I informed the
respondents, when first contacted via email or phone regarding the possibility to hold an interview, that this was would be conducted for the purposes of my doctoral study on a specific topic and was not a meeting to address work duties. At the beginning of each interview, the respondents were provided with the description and goals of the present research. All study participants filled in the informed consent (presented in Appendix E) on paper for offline interviews and electronically for online meetings.
Analysis bias happens when the researcher’s beliefs lead to specific interpretation and confirm their personal experience.
The fact that data were analyzed only by one person led to the risks of:
inaccurate transcription of interviews
As I indicated in section 3.7, the first round of transcription was done by professional
transcribers. Then, in order to minimize the bias, I double-checked each interview in order to verify the accuracy of the transcription.
missing any ideas during data analysis
I fact-checked the analysis with my former UrFU colleagues. First, I received four data sets on factual information about international students, international faculty members and research published on international databases by UrFU colleagues upon the approval of the Rector, which I compared with data in official documents and the website. Second, I fact-checked the conclusions on the analysis of context and RPE with an UrFU colleague, who was responsible for data monitoring at the university. I also did one more round of critical review of my coding results in Atlas.TI software and an additional round of themes revision themes after receiving the comments of my research supervisors.
It is also important that researchers interpret data within the framework of qualitative studies and not present their conclusions as the one and only true picture:
Principally, … our challenge is not to try and convince that qualitative work reflects objective, opinion-free neutrality. Rather, it is to better articulate the unique value
that qualitatively derived knowledge can play within a system that measures impact through an evidence-based decision-making lens. (Galdas, 2017)
Social scientists agree that qualitative research, in general, and qualitative interviews, in particular, cannot be totally objective. Janesick (1998, p. 48) noted that “qualitative
research depends on the presentation of solid descriptive data” and the interpretation of data depends on the perspective of any concrete researcher. Edwards and Holland (2013, p. 84) wrote that “by virtue of being human, researchers are not neutral and objective enquirers in qualitative interviews but are emotionally engaged participants who are sharing an experience with the interviewee”. This viewpoint is aligned with my social constructivist position, described in section 3.2, since “social constructivism tells us we build knowledge as ways of understanding the world, and that these ways of understanding are a subset of how the world could be understood” (Jackson, 2010).
Moreover, several authors noted that such biases in qualitative research are not only inevitable, but also desirable (Glaser, 1992; Strauss & Corbin, 1998). As regards case studies, Wolcott (1991) argued about the impossibility of predicting a single correct interpretation of a case that is inapplicable to this research method. Thus, in conducting the present research, I was actively searching for ways to take a more neutral stance and I was always aware of how my insider experience might influence the interpretation.