IEEE Consumer Electronics Magazine - July/August 2023 - 13

diagnose traumatic brain injury and concussions.7
As of 2022, more than 300 AI/ML-based
healthcare devices have been approved by the
FDA, and there are many more in the development
pipeline.8 Many direct to consumer AI
devices for health monitoring and well-being
are already available and in use by patients
worldwide.9
AI is thus actively shaping new " smart " healthcare
products, with the ability to learn complex
trends from large datasets, and continually
improve their core algorithms with new data over
the lifecycle of the product. Many innovative AIdriven
technologies are beginning to emerge,
such as virtual health assistants, personalized
therapeutics, smart digital pills, and remote
robotic surgeries on the horizon. However, AI is
prone to misuse when deployed inappropriately
(e.g., because of over-trust), or disuse when it
should be used but is not (e.g., because of undertrust).10
AI can be abused when implemented
without due regard for the medical practitioners'
capabilities or the patients' interests.11 AI algorithms
may also be unintentionally biased in their
diagnosis due to biases in their training datasets
or produce outcomes that are difficult for experts
to comprehend. AI in smart healthcare thus raises
many new ethical concerns related to transparency,
bias, responsibility, safety, privacy, justice,
and autonomy, as will be discussed later.
The term ethics refers to the area of philosophy
concerned with the study of what is right
and wrong. The subfield of normative ethics
focuses on theories that provide general moral
rules governing our behavior.12 There are three
popular categories of theories-deontology, teleology
(consequentialism), axiology (virtue
ethics)-that are relevant for resolving ethical
dilemmas, such as those that accompany the use
of AI-driven healthcare products. Deontology
makes use of rules and processes to guide ethical
decisions. These rules imply that acts must be
performed regardless of their consequences, e.g.,
the truth should always be told, no matter what
its consequence. In contrast, consequentialism
emphasizes outcomes rather than the process. It
is an outcome-based approach that defines ethical
behavior based on consequences (good or
bad). Ethical decisions are the ones that create
the most good. Utilitarianism is a popular
July/August 2023
consequentialist approach that aims for the maximum
good for the most number of individuals.
Finally, virtue ethics is guided by the question, " If I
was a good person, what would I do? " Key to this
theory is a reliance on core ethical virtues, such
as justice, moderation, loyalty, compassion, and
courage to guide ethical actions. The focus in virtue
ethics is therefore on virtuous personal character
when it comes to making decisions rather
than universal rules or consequences. These theories
are widely reflected in different spheres of
aspirational modern societies and in the codes
and values of professional institutions.
DIFFERENCES IN MEDICAL AND
AI ETHICS
Practitioners in the medical domain are wellversed
with the four principles of clinical
biomedical ethics: justice (fair, equitable, and
appropriate treatment of patients), respect for
autonomy (relating to a patient's capacity to selfdetermine),
beneficence (obligation to act for the
benefit of the patient), and nonmaleficence
(avoiding harm to the patient).13 Principlism in
medicine represents a theoretical moral framework
that combines traditional ethical standards
with the requirements of practitioners, research
ethics committees, and medical institutions for
practical ethical decision-making.14 While this
ethical principlism provides guidance for clinical
decision-making and setting health policy,15 there
are four important differences between medicine
and AI development that suggest that a principled
approach for AI development (for healthcare and
other contexts) may have limited impact on its
design and governance. Compared to medicine,
AI development lacks:
1) Common goals and fiduciary duties: While medicine
is broadly guided by the common goal of
promoting thehealth andwell-beingof patients,
AI development for healthcare is driven by the
private technology sector where there are constant
pressures to maximize profit, reduce
costs, and make decisions that prioritize the
interests of the company above all else. In medicine,
as in other formal professions, regulatory
frameworks and professional codes have
established fiduciary duties towards patients,
which are facilitated by common goals and
13
IEEE Consumer Electronics Magazine - July/August 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - July/August 2023
