IEEE Technology and Society Magazine - March 2021 - 50

AI for a Fair, Just, and Equitable World

The secondary role to which the document relegates participation of the wider society, both in
AI development policies and in the AI regulatory
efforts, is disturbing.

R&D in ethical AI
The EU AI development strategy should seek to
promote independent and scientifically rigorous
research in the ethical aspects of AI applications,
both theoretical and practical/empirical perspectives, from philosophical reflection to the development of risk-mitigation strategies, impact assessment
frameworks, and applied ethics method and tools.
Such methods and tools are not meant to replace
ethics legislation and good practice manuals, but
rather to support their implementation. Both for the
development of impact assessment frameworks and
for that of the applied ethics methods and tools, a
multidisciplinary approach is key to their utility.
The SDGs themselves define the basis of an impact
assessment framework, defining indicators that
cover multiple perspectives, including social costs
and benefits, which can take into consideration the
cultural values of the communities concerned.
Theoretical and empirical research on ethical AI
lags far behind purely technological AI research, for
which reason, the former should be given strategic
priority over the latter. It is imperative to accelerate
the development of AI ethical assessment frameworks to guide the future of AI in Europe. A commitment from the private sector is required to prioritize
the assessment of algorithms and platforms.

Encouraging multiscale
experimentation
Support tools for the application of ethical
principles developed in research laboratories are
currently at an early state of research so that their
usefulness and impact are yet to be demonstrated.
Examples of quantified good practice in the application of applied-ethics tools in real-world projects
are needed. The promotion of R&D projects for the
experimentation of AI with an ethical approach
involving industry, the public sector, and research
centers, would help create such good practices.
Applied ethics tools also exist in the private sector and are used in internal consultancy. In general,
these tools are in a very immature state and are not
ready to be adopted by professionals in the short
term, although in some very specific fields, effective

50

tools with limited scope may already have been
developed. Moreover, for the time being, using such
tools does not provide any competitive advantage,
which is why it is important to encourage them
through proactive policies that promote their incorporation into business models. For impartiality,
AI ethics assessment should be outsourced in the
same way that many assessment, certification, and
quality control processes are outsourced. For the
assessment process to be homogeneous requires
standards at national and European levels to guide
and regulate the assessment tasks.

Conflicts between AI and human rights
protection
The document refers to the defense of human
rights, but only develops the arguments with regard
to civil and political rights (privacy, political rights,
and freedoms), ignoring social, economic, and
cultural rights. This is closely related to concerns
about the inequality that the massive use of AI could
generate.
To prevent Al-based technological developments
from causing human rights violations, the system of
accountability needs to be restructured.
On many occasions, the end user is assigned the
responsibility of verifying the guarantees of noninfringement of rights such as data privacy. However,
the user lacks the capacity and vision to analyze
complex systems. It is, therefore, necessary to create
environments for public debate, where dialogues
are held between researchers, developers, and
members of civil society, as stated in the Zaragoza
Declaration [35].
Assuming the need for transparent governance of
AI, the areas of AI research to be prioritized should
include research on verifying the fairness of algorithms, with a view to integrating such verification
mechanisms in a certification process to be carried
out by an independent authority.
Quintanilla [36] coined the term " endearing
technologies, " as opposed to " alienating technologies " ; this idea being consistent with the responsible
research paradigm. We believe that such technologies are more respectful of fundamental rights. Some
of their characteristics, as defined by Quintanilla, are:
openness (the software is free); versatility (it allows
alternative uses); docility (its operation and control
depends on a human operator); no planned obsolescence (repair is promoted more than replacement);

IEEE Technology and Society Magazine
IEEE Technology and Society Magazine - March 2021

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