IEEE Technology and Society Magazine - March 2021 - 49

is beyond the capability of most private companies.
Furthermore, fragmented development of AI impact
assessment by many different companies could
increase biases and, at the same time, be a drain
on resources of smaller companies, making them
uncompetitive against the platforms of the digital
giants [27].

Avoiding ethnocentricity and
androcentricity
Although the need to avoid ethnic and gender
discrimination and biases in Data Science projects
is mentioned in the document (guaranteeing the
representation of all groups in the data), there is no
mention of the need to design applications adapted
to different cultural sensitivities and values (culture-aware systems, see [28]) or the need to enhance
the representation of all groups in the research,
design, and development of AI applications (through
mechanisms such as community-based development
methodologies, citizen science, etc.). We emphasize
the multicultural strength of Europe and the opportunity that this represents for AI, as well as the synergies
that can exist to build a more cohesive Europe if an
AI sensitive to people and to these cultures is developed. This is only possible with the strengthening of
pan-European research involving both the public and
private sectors.
The document does speak of the need to promote women's vocations in AI technologies, but not
to encourage female entrepreneurship in AI. In the
computer industry, only 1%-2% of new companies
receiving venture capital funds are run by women,
even though women-led companies get a 200%
return on investment [29]. This is the area where
gender biases in the technology industry are most
apparent. This is a missed opportunity to develop a
more inclusive and innovative AI.

Universalizing the benefits
The White Paper should position itself with
respect to " open science, " in the sense of the upcoming UNESCO Open Science Recommendation [30],
or " open AI " paradigm (in particular, open source
and open algorithms to facilitate the dissemination
of the benefits of AI). A paradigm of knowledge sharing, replicability, and openness is needed. Technology is not always neutral (see [31] and [32]) and
design much less so (see [33]), and, furthermore,
the adoption of AI is not within the reach of all.

March 2021

For these reasons, AI will not diminish inequalities
on its own unless specific policy lines are developed with these objectives in mind, along with the
frameworks for ethical design, implementation, and
impact assessment needed to support them. There is
a risk of AI creating new inequalities as well as greater
difficulties in the process of improving opportunities
for vulnerable groups, thus creating an additional
barrier between those who can benefit from AI and
those who cannot. The risk is of creating a new AI
gap, which would overlap with the already worrying
digital divide that exists globally today. It would also
be highly desirable to mention the promotion of a
social economy in the AI sector.

Education and digital literacy
In our opinion, more emphasis should be placed
on the need for multidisciplinary education in AI, for
raising the ethical and social awareness of AI professionals and researchers and for the education of
society, in general, regarding digitization and AI, one
important reason being to raise awareness of its ethical and sociopolitical implications.
As recognized in several recent reports such as
[34], ethics should be embedded in the engineering professional practice and education, providing
students and engineers with the tools necessary to
build an ethical orientation into their inventions.
Therefore, regarding skills, educational institutions
(especially universities) have a responsibility to provide curricula with the necessary multidisciplinarity,
including a strong ethics-education component that
incorporates philosophical, social, political, psychological, and risk analysis aspects of the development
and implementation of AI technologies. Pedagogical
research in these aspects should be promoted within
the EU AI strategy.

Dialogue between all social actors
Dialogue between government and industry
about the use of AI technologies is welcome but
insufficient; any such dialogue should also involve
the different social actors (third sector, civil society, academia,.etc.). To facilitate participation, it is
essential that this dialogue be carried out in a transparent fashion, particularly as far as defining a regulatory framework for AI is concerned. The social
value of AI is determined collectively; the wider the
participation, the greater the confidence in, and the
social appropriation of, the technology.

49
IEEE Technology and Society Magazine - March 2021

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