IEEE Technology and Society Magazine - March 2021 - 47

to dystopic scenarios based on control and surveillance [11]. The competition-oriented view is also
scientifically reductionist, in that actions and policies tend to techno-optimism and technological
" solutionism " rather than being informed by deep
and integrative research on issues such as avoiding
" bias-by-design. "
Inclusiveness is a key element of a humanbased AI, being also the motivation of the data revolution for sustainable development promoted by
the UN in [3]. AI should help empower vulnerable
groups, of which Europe also has its share, rather
than contribute to marginalizing them. There is
a significant risk of AI increasing socioeconomic
disparities, notably by extending the existing digital divide into an AI divide (e.g., built-in social
biases increasing discrimination against minorities, infrastructure poverty creating competitive
disadvantages) [12], [13]; a competition-oriented
AI strategy will make this risk more acute and will
likely also have severe negative effects on the
environment [10]. Universalizing the benefits of AI
requires a strategy in which openness, in the sense
of the upcoming United Nations Educational, Scientific and Cultural Organization (UNESCO) Open
Science Recommendation [14] (open source,
open data, open algorithms, etc.), can play an
important role as an enabling factor, in particular,
facilitating dissemination and auditing.
In the last few years, multiple ethical codes
and guidelines for AI development and deployment have been drawn up by many different
organizations, mostly in response to growing
awareness of possible adverse effects of AI.
However, the resulting fragmented and largely
corporate-based mosaic of ethical frameworks,
codes, and guidelines, many of which suffer
from deficiencies such a lack of scientific rigor,
subjectiveness, incoherence, superficiality, and
redundance, creates confusion [15], [16]. Moreover, the dominant reductionist vision promotes
accountability for a number of commitments to
technological principles, such as transparency,
but not to upholding values, such as equity [17],
[18]. Policy documents related to ethics and AI
have also focused more on ethical frameworks
than on possibilities for enhanced regulation
[19]. However, there is a perceived need for
new regulatory frameworks to reinforce the recommendations of policy documents, rather than

March 2021

relying on self-regulation and individual commitments, as has so far been the case1 purportedly
due to fears that regulatory intervention could
limit potential innovative capacity. The objective
of an inclusive, sustainable, and human-centered
AI in Europe will likely require a normative framework at the European level. Financial and regulatory stimuli are required to foster SDG-driven AI
and public-private collaboration in the sharing
of technology and data [22].
Furthermore, a human-centered AI should be
human rights-based [23]. Although there has been
some limited discussion at the European government level of the impact of AI on human rights,
especially regarding the right to privacy, the impact
on social, economic, and cultural rights has so far
received little attention.
Data quality is of particular importance for AI
applications, dictating the need for mechanisms and
metrics to safeguard this quality, especially to ensure
that training data does not result in bias. Open
access to public-sector data sources with high scientific and social value is one way of contributing to
reducing such bias. Ensuring the use of quality data
also needs to address the thorny issue of public and
third sector access to any private-sector data that is
of vital or significant public interest. Increasing and
optimizing access to, and usage of, data inevitably
involves the interconnection of diverse data sources,
bringing to the fore issues of privacy and security,
which need further investigation.
Reliable and trustworthy AI understood in merely
technical terms is not sufficient. The evaluation of
impact, grounded in both the analytical and the
social sciences, is essential to a human-rights-based
approach [24]. Experimentation of AI systems at different scales prior to deployment is of great importance, and impact monitoring of such systems after
deployment is crucial [16]. More (reproducible)
research is needed on impact indicators for AI systems [25], on the validation and certification of
properties such as explainability and algorithmic
justice and on the development of applied-ethics
1
Businesses should provide remediation for any adverse human rights impacts
caused by AI through legitimate mechanisms. Although the International Law
of Human Rights establishes the duty of the state to provide effective remedies
through judicial and other mechanisms to those who have suffered business-related
human rights abuses, it rarely happens. The UN Guiding Principles on Business and
Human Rights can help, as they can clarify how preexisting international human
rights standards apply to business activities, and provide useful guidance on how
businesses can operate in a rights-respecting manner [20], Why Embracing Human
Rights Will Ensure AI Works for All [21].

47
IEEE Technology and Society Magazine - March 2021

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