IEEE Technology and Society Magazine - September 2023 - 84

Feature
Figure 6. Overall percentage of proposed
practices in each principle.
are identified across the datasets and there are a certain
number of discussions involving the inclusiveness
and equity of access to AI technology. However,
the trend of calling for future standards and practices
remains relatively low compared to other principles.
A great number of discussions revolve around ethical
challenges in technical aspects with focuses on
sampling bias within training data of AI and lack of
demographic representation in data collection [31].
Proposals regarding such ethical issues point more
toward categories such as design and deployment
of AI, leaving a gap between the growing inequalities
of technology resources distribution and actual
attention granted toward resource-scarce regions and
populations. In addition to that, data collection and
relevant research are also limited in regions suffering
from insufficient accessibility to technology, making
it hard for AI developers and regulators to bring up
discussions and future proposals targeting the specific
challenges these regions face.
Three themes
By using a GT framework to analyze policy documents
on AI ethics in healthcare, the study was able
to generate theoretical coding that provides insights
into the key ethical considerations surrounding the
development and deployment of AI in healthcare.
In conclusion, our content analysis of various
policy documents on AI ethics in healthcare reveals
several important findings. First, the role of designers,
developers, and administrators in the design
of AI systems is a crucial consideration that should
be given priority. Without their active involvement
in the development process, ethical considerations
may be overlooked, resulting in unintended consequences
that could potentially harm patients.
84
Second, our analysis highlights the importance
of inclusiveness and equity in the development and
deployment of AI in healthcare. These principles are
not only essential for ensuring that the benefits of AI
are accessible to all, but they also interact with other
ethical considerations, such as privacy and autonomy.
Finally, we note that there is a relative insufficiency
in the current discussion regarding sustainability
and energy efficiency in the development and
deployment of AI in healthcare. While these principles
may not be as frequently mentioned as others,
they are nonetheless important considerations that
should be addressed to ensure that the long-term
impact of AI on the environment is minimal.
Overall, our analysis suggests that while there is
a growing awareness of the ethical considerations
surrounding AI in healthcare, there are still areas
that require further attention and discussion. By
actively engaging with these issues, policymakers,
developers, and administrators can ensure that AI
is developed and deployed in a way that is ethical,
equitable, and sustainable.
Limitations
By extending the data pool to a global scale, the
study was able to build a data pool resulting in more
than 10,000 of coded segments for analysis. However,
it has also been noticed in the literature search
process that discussions spanning all three fields of
AI technology, healthcare, and ethical regulations
could be hard to find. Thus, the study had to incorporate
numerous documents that only addressed
two out of the three topics. Unintentional biases
have been imposed as a result. For instance, healthcare-related
search terms such as " Human Control
Over Medical Decisions " would not be reflected as
much in documents that only discussed general AI
applications. The methodology of constructing the
coding system also contains some limitations. The
search terms are generated with references to relevant
works while also incorporating ideas from web
databases and expertise from the field. The lack of a
rigid standard might lead to the inability to capture
complete context and information. The structure of
code labels also limits the capabilities of differentiating
between the frequency and relevance of coded
segments. For instance, a keyword that appears in
a topic sentence might not imply the same significance
as its appearance in a description of a realworld
incident.
IEEE Technology and Society Magazine
IEEE Technology and Society Magazine - September 2023

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