IEEE Technology and Society Magazine - September 2023 - 79

major intergovernmental organizations were identified
as eligible sources.
Therefore, based on the above inclusion criteria,
one government document, four organization documents,
and two academic articles are selected.
After further selection processes based on theoretical
sampling technique [23] and selective coding,
the data pool was eventually extended to ten government
and organizational documents and five
academic articles, which were the final number of
sources used for content analysis in the next step.
Figure 1 summarizes the procedure of selecting and
retaining relevant and suitable articles, following the
preferred reporting items for systematic reviews and
meta-analyses (PRISMA) framework [24].
In the next step, content analysis is performed
utilizing these documents as datasets. The six key
ethical principles proposed by Ethics and Governance
of Artificial Intelligence for Health by WHO ( " the
WHO document " ) are used as the basic structure of
the content analysis. Under each principle, a set of
subprinciples are established to accurately extract
essential content from data sources. The coding
function of MAXQDA is used as a major tool. Comprehensive
guidance has been introduced on using
MAXQDA for qualitative analysis by reputable books
to ensure that researchers have access to valuable
resources. For example, Woolf and Silver's book
[25] introduces the five-level QDA™ method, providing
researchers with a structured approach to
conducting qualitative analysis using MAXQDA.
It covers various aspects of data analysis, including
coding and advanced techniques, making it an
essential resource for effectively utilizing MAXQDA
in qualitative studies. Kuckartz and Rädiker [26]
offer a comprehensive guide to analyzing qualitative
data in MAXQDA, encompassing different data
formats such as text, audio, and video. It presents a
step-by-step approach to coding, querying, and visualizing
data, making it an invaluable resource for
researchers seeking to leverage MAXQDA effectively
in their qualitative research. Silver and Lewins [27]
provide a broader perspective on using software in
qualitative research, including a dedicated section
on MAXQDA. It covers the fundamentals of using
MAXQDA for data management, coding, and analysis,
along with integrating software into the qualitative
research process. The practical examples and
step-by-step guidance make it an accessible and reliable
resource for researchers utilizing MAXQDA.
September 2023
Figure 1. Article selection and retention
process.
Code labels generated by subprinciples are input
into the autocode search engine to find out occurrences
of each subprinciple in different documents.
Due to limitations of the autocode function, manual
checks are also performed to ensure the accuracy of
coded segments.
Coding and data analysis
Adapted from Strauss and Corbin's [23] model
of GT, multiple rounds of coding and analysis were
conducted with constant comparison between the
documents, the research interests, and the existing
literature. As demonstrated by [28], the coding process
guided by the GT framework is emergent and
recursive, sometimes unstructured. The essential
premise is that the process guided by GT is varied
among different studies and subject to interpretations
by different researchers [29]. Accordingly, our
interpretation of primary coding segments and key
coding methods have been constantly adjusted over
the course of the research. Hence, to best depict the
coding and analysis activities, we are reconstructing
it as much as we are reporting on them.
Six ethical principles endorsed by the WHO
are used as the backbone of the coding system, as
shown in rectangles in Figure 2. The principles represent
major directions of AI development in existing
79
IEEE Technology and Society Magazine - September 2023

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - September 2023
