IEEE Technology and Society Magazine - June 2021 - 57

* Right to equality and nondiscrimination: Applications
that use AI, and especially ML, must be
" trained " using data. Where that data incorporates
unfairness, such as discrimination, this can
be replicated in the new application. Where
training data is collected and used well, new
technologies such as AI can enable better service
delivery, especially for vulnerable groups.
Unequal access to critical new technologies can
exacerbate inequalities, especially where access
is affected by factors such as socioeconomic status,
disability, age, or geographical location.
* Accessibility: New technologies can provide
new ways to deliver services, thereby increasing
accessibility for people with disability and others.
Reduced cost of services through affordability of
new technology can promote equality for people
with disability by ensuring progressive realization
is achieved faster and reasonable adjustments are
more affordable. New technologies can increase
barriers for people with disability if they are used
in products and services in ways that are not
accessible.
* Right to privacy: The ease of collecting and using
personal information through new technologies
such as facial recognition can limit the right to
privacy. Personal data can flow easily and quickly,
across national and other borders. This can make
privacy regulation and enforcement more difficult.
It can be difficult to " correct " or remove personal
information once communicated. The ease
of communicating and distorting personal information
(e.g., through " deep fakes " ) can lead to reputational
damage and other harms.
Very useful initiatives in this sense are the Toronto
Declaration on Protecting the Rights to Equality and
Nondiscrimination in Machine Learning Systems
[13] (May 16, 2018) and the strategic article called
" Artificial Intelligence: Human Rights & Foreign Policy
Implications " [14].
Principles inspiring the new regulation
From a rights-based approach, principles are
essential. Human rights are universal and interdependent;
participation is crucial; we cannot forget
nondiscrimination, participation, and accountability.
Each of these principles must be in the center of our
June 2021
technological projects.1 A human-rights approach is
not formally explicit in the EU White Paper on artificial
intelligence regulation, although there are some
mentions to the citizen rights through a number of
examples. A human-rights approach is based on the
principle of universality and should be the framework
for a regulation that has the aspiration of being the
first global mover laying the foundations for AI regulation.
Not everybody in different parts of the world
or even within the same community have the same
vision of the best, the more trustworthy, or ethical digital
future. Instead, a human-rights approach would
rely on the needed international consensus to start
building a shared vision around AI regulation [11].
We, therefore, challenge the premise that building
trust should be the base of regulation in light of the
human-rights approach. The White Paper, in the beginning
of Chapter 5, " An ecosystem of trust: Regulatory
framework for AI " states that " Citizens fear being left
powerless in defending their rights and safety " and
proposes regulation as the solution for this fear or distrust.
But distrust is a healthy societal mechanism when
rights could be threatened. Trust is not the objective,
neither is the uptake of AI technology. The objective is
the individual and community safety and well-being in
a broad sense. Other objectives, like competitiveness
of European AI companies, should be clearly subordinated
to the former. Following Ryan [15], we also think
that establishing a relationship of trust with AI is a trust
misplacement in a device (like trusting a company, for
this sake) and is a form of reliance, more than trust.
The EU White Paper concentrates on the protection
of individual rights. We also advocate for an
explicit inclusion of public and common goods and
rights that the new regulatory framework should also
protect. Public goods such as reliable public information,
stable political system, fair economic system
with balanced powers of different stakeholders and,
in general, all conditions for our democracy, and the
rule of law in our society should be considered for
protection apart from individual rights.
The framework used by the authors of the EU White
Paper seems to be based on a stakeholder vision
of attending everyone's interests, which is a broadly
accepted approach, but there is some indication that
the White Paper is somewhat skewed toward the interest
1An example of how to include human rights principles in technology could
be the following: If there is an AI system which is going to be introduced in the
market, but it lacks accessibility-it is not accessible for deaf people, for
example-the AI technology is not rights based, as it is against the
nondiscrimination principle. We should rethink the project and include the
electronic accessibility.
57
IEEE Technology and Society Magazine - June 2021

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - June 2021
