Chapter 8

Conclusion

Abstract The conclusion briefly summarises the main arguments of the book. It
focuses on the requirements for mitigation options to be used to address the ethical
and human rights concerns of artificial intelligence. It also provides a high-level
overview of the main recommendations brought forth in the book. It thereby shows
how conceptual and empirical insights into the nature of AI, the ethical issues thus
raised and the mitigation strategies currently being discussed can be used to develop
practically relevant conclusions. These conclusions and recommendations help to
ensure that AI ecosystems are developed and shaped in ways that are conducive to
human flourishing.

Keywords Requirements for AI ethics · Recommendations for AI · AI

governance · Ethics and AI ecosystems · AI regulation

Technology is part of human life. Its development and use have the potential to raise
ethical concerns and issues – and this will not change. Ethics, understood as our
struggle to determine what is right and wrong and our reflection on how and why we
make such a distinction, is not subject to resolution. While we may agree on what
is right and wrong in many cases, this agreement is always partial, temporary and
subject to revision. We may, however, be able to agree on some general and abstract
principles. In this book I have suggested that human flourishing is such a principle.
If we agree on that, then we can think through what the application of the principle to
a technology such as AI can mean. This exercise can help us understand the specific
issues that arise, why they arise and how we can evaluate them. It can also help us
think through what we can do about them, and may even help us resolve some of
them to universal satisfaction.
Several aspects that I have focused on in this book can, I hope, make a novel
and interesting contribution to the AI and ethics debate. I started by looking at the
concept of AI. “Artificial intelligence” is not an innocent and morally neutral term.
It is emotive because it points to a characteristic of humans (and to some degree of
other animals) while implying that this characteristic can be artificially replicated.
This implication has consequences for how we as humans see ourselves and our role

© The Author(s) 2021 117

B. C. Stahl, Artificial Intelligence for a Better Future,
SpringerBriefs in Research and Innovation Governance,
https://doi.org/10.1007/978-3-030-69978-9_8
118 8 Conclusion

in the world. Artificial intelligence is also often contrasted with human intelligence,
implicitly suggesting or explicitly asserting that machines can or even should replace
humans. Again, this touches deeply rooted views of what humans are.
In order to render the discussion more accessible, I have proposed a new cate-
gorisation of the AI debate. My suggestion is that we distinguish between three
perspectives on AI: machine learning or narrow AI, general AI and converging
socio-technical systems. These three perspectives on the technology are enlightening
because they align with the categorisation of ethical issues on AI: first, ethical issues
related to machine learning; second, general issues related to living in a digital world;
and third, metaphysical issues posed by AI. These distinctions thus provide a better
understanding and overview of AI ethics in a very busy and often overwhelming
public and academic debate.
While these categorisations clarify the debate, they say very little about what could
or should be done about the issues. One of the problems in this type of normative
discussion is that it is unclear how recommendations or prescriptions can be justified.
On what grounds could we say that technical applications should be developed,
promoted, avoided or prohibited? Drawing on the idea of human flourishing allows a
normative point of reference to be established that is consistent and compatible with
the main ethical theories and can provide a framework for thinking about normative
questions without presupposing substantive moral positions.
The idea of human flourishing has the added advantage of not requiring a strict
distinction between ethics and law, both of which are normative constructs that could
promote or inhibit flourishing. This is particularly important in light of the numerous
existing legal and regulatory rules that already guide the development and use of
technology, including AI.
Drawing on rich empirical work, I analysed ethical concerns and suggested inter-
ventions, mitigations and governance approaches to promote the benefits of AI and
avoid or address its downsides.
One problem in the AI ethics discussion is its high level of complexity. Any
attempt to match individual issues with stakeholders and mitigation options runs
into several problems. First, the number of possible combinations of stakeholders,
mitigations and ethical issues to be addressed is such that it is impractical to try
to understand the field using such a straightforward approach. Second, and more
important, the different components of the interaction are not independent, and an
intervention in one part is likely to have consequences in another part. As this type
of dynamic relationship lends itself to being described using a systems perspective,
I have adopted the now widely used ecosystem metaphor and applied it to the AI
discourse.
The question of what needs to be done to ensure that AI ecosystems are conducive
to human flourishing was then tackled through the ecosystem metaphor. This led me
to investigate, from an ethical perspective, the implications of using the ecosystem
metaphor, a question that is not yet widely pursued in the AI field. In addition, I
analysed the challenges that the ecosystem approach to AI and ethics raises and
the requirements that any intervention would need to fulfil, and I concluded with
suggestions to take the debate further and provide input into discussions.
8 Conclusion 119

The analysis pointed to three groups of requirements that interventions into AI

ecosystems need to fulfil, in order to increase their chances of successfully promoting
human flourishing:
• Interventions need to clearly delineate the boundaries of the ecosystem:
Systems boundaries are not necessarily clear and obvious. In order to support
AI ecosystems, the boundaries of the ecosystem in question need to be clearly
located. This refers not only to geographical and jurisdictional boundaries, but
also to conceptual ones, i.e. the question of which concept of AI is the target
of intervention and which ethical and normative concepts are at the centre of
attention.
• Interventions need to develop, support, maintain and disseminate knowledge:
The members of AI ecosystems require knowledge, if they are to work together to
identify ethically desirable future states and find ways of working towards those.
AI as a set of advanced technologies requires extensive subject expertise in the
technologies, their capacities and uses. In addition, AI ecosystems for human
flourishing require knowledge about concepts and processes that support and
underpin ethical reflections. And, finally, AI ecosystems need mechanisms that
allow for these various bodies of knowledge to be updated and made available to
members of those ecosystems who need them in a particular situation.
• Interventions need to be adaptive, flexible and able to learn: The fast-moving
nature of AI-related innovation and technology development, but also of social
structures and preferences as well as adjacent innovation ecosystems, means that
any intervention into the AI ecosystem needs to incorporate the possibility and,
indeed, likelihood of change. Governance structures therefore need to be flexible
and adaptable. They need to be open to learning and revisions. They need to be
cognisant of existing responsibilities and must build and shape these to develop
the ecosystem in the direction of human flourishing.
These requirements are deduced from the nature and characteristics of AI inno-
vation ecosystems. They are likely to have different weights in different circum-
stances and may need to be supplemented by additional requirements. They consti-
tute the basis of the recommendations developed in this book. Before I return to these
recommendations it is worth reflecting on future work.
The work described in this book calls for development in several directions.
An immediate starting point is a better empirical understanding of the impact of
AI and digital technologies across several fields and application areas. We need
detailed understanding of the use of technologies in various domains and the conse-
quences arising. We also need a much broader geographical coverage to ensure that
the specifics of different nations, regions and cultures are properly understood.
Such empirical social science research should be integrated into the scientific
and technical research and development activities in the AI field. We need a strong
knowledge base to help stakeholders understand how particular technologies are
used in different areas, which can help technical researchers and developers as well
as users, deployers, policymakers and regulators.
120 8 Conclusion

The insights developed this way will need to be curated and made available to
stakeholders in a suitable way. To a large extent this can be done through existing
structures, notably the scientific publication process. However, issues of legislation,
regulation and compliance require special gatekeepers who can lay claim not only
to a high level of scientific and technical expertise, but also to normative legitimacy.
The idea is not to install a tyranny of the regulator, but to establish ways that help
stakeholders navigate the complexity of the debate and spaces in which organisations
and societies can conduct a fruitful debate about desirable futures and the role that
technologies should play in them.
The discussion of the ethics of AI remains high-profile. Numerous policy and
regulatory proposals are likely to be implemented soon. The causes of the high
level of attention that AI receives remain pertinent. The technologies that constitute
AI continue to develop rapidly and are expected to have a significant social and
economic impact. They promise immense benefits and simultaneously raise deep
concerns. Striking an appropriate balance between benefits and risks calls for difficult
decisions drawing on expertise in technical, legal, ethical, social, economic and other
fields.
In this book I have made suggestions on how to think about these questions and
how to navigate the complexity of the debate, and I have provided some suggestions
on what should be done to facilitate this discussion. These recommendations have the
purpose of moving AI ecosystems in the direction of human flourishing. They satisfy
the three requirements listed above, namely to delineate the ecosystems boundaries,
to establish and maintain the required knowledge base and to provide flexible and
adaptive governance structures. In slightly more detail (see Chapter 7 for the full
account), the recommendations are:
• Conceptual clarification: Move beyond AI (7.3.1)
The concept of AI is complex and multi-faceted (see Chapter 2). The extent of the
ecosystems concerned and the ethical and human rights issues that are relevant in
them depend to a large degree on the meaning of the term “artificial intelligence”.
Any practical intervention should therefore be clear on the meaning of the concept.
It will often be appropriate to use a more specific term, such as “machine learning”
or “neural network”, where the issues are related to the characteristics of the
technology. It may also be appropriate to use a wider term such as “emerging
digital technologies”, where broad societal implications are of interest.
• Excellence and flourishing: Recognise their interdependence (7.3.2)
In the current discussion of AI, including some of the policy-oriented discourses,
there is a tendency to distinguish between the technical side of AI, in which
scientific and technical expertise is a priority, and the ethical and human rights
side. This blurs the boundaries of what is or should be of relevance in an AI
ecosystem. The recommendation points to the fact that scientific and technical
excellence must explicitly include social and ethical aspects. Work on AI systems
that ignores social and ethical consequences cannot be considered excellent.
8 Conclusion 121

• Measurements of flourishing: Understanding expected impacts (7.3.3)

In order to react appropriately to the development, deployment and use of AI, we
must be able to understand the impact they can be expected to have. It is therefore
important to build a knowledge base that allows us to measure (not necessarily
using quantitative metrics) the impact across the range of AI technologies and
application areas. While it is unlikely to be possible to comprehensively measure
all possible ethical, social and human rights impacts, there are families of measure-
ments of aspects of human flourishing that can be applied to AI, and these need
to be developed and promoted.
• AI benefits, risks and capabilities: Communication, knowledge and capacity
building (7.3.4)
The knowledge base of AI ecosystems needs to cover the technical side of AI
technologies, to ensure that the risks and potential benefits of these technologies
can be clearly understood. This knowledge, combined with the measures of human
flourishing in the preceding recommendation, is required for a measured view
of the impact of AI systems and a measured evaluation of their benefits and
downsides. This knowledge base that AI ecosystems must be able to draw on,
in order to make justifiable decisions on AI, is dynamic and can be expected to
evolve quickly. It therefore needs to develop mechanisms for the regular updating
and development of expertise and means of disseminating it to those who need it.
• Stakeholder engagement: Understanding societal preferences (7.3.5)
The broad and all-encompassing nature of AI and its possible impacts means
that decisions shaping the development, deployment and use of AI and hence
its societal impact must be subject to public debate. Established mechanisms of
representative democracy have an important role to play in guiding AI governance.
However, the dynamic and complex nature of the field means that additional
mechanisms for understanding the views and perceptions of stakeholders should
be employed. Involving stakeholders in meaningful two-way communication with
researchers, scientists and industry has the advantage of increasing the knowledge
base that technical experts can draw on, as well as improving the legitimacy of
decisions and policies resulting from such stakeholder engagements.
• Responsibility for regulation and enforcement: Defining the central node(s)
of AI ecosystems (7.3.6)
AI ecosystems do not develop in a vacuum but emerge from existing technical,
social, legal and political ecosystems. These ecosystems have developed a plethora
of mechanisms to attribute responsibility with a view to ensuring that the risks and
benefits of emerging technologies are ascribed appropriately. The emergence of
AI ecosystems within these existing environments means that existing roles and
responsibilities need to be suitably modified and developed. This calls for a way
of coordinating the transition to AI ecosystems and integrating them into estab-
lished contexts. The shifting networks of responsibilities that govern emerging
technologies will therefore need to evolve ways of developing formal and informal
governance structures and monitoring their implementation. This calls for the
establishment of central nodes (e.g. regulators, agencies, centres of excellence)
122 8 Conclusion

that link, guide and oversee AI ecosystems, and relevant knowledge and structures
to ensure the technologies contribute to human flourishing.
I hope that this book and the recommendations that arise from it help strengthen
the debate on AI and ethics. The book aims to support the appropriate shaping of
AI ecosystems. In addition, its message should reach beyond the current focus on
AI and help to develop our thinking on the technologies that will succeed AI at the
centre of public attention.
Humans are and will remain tool-using animals. The importance of technical tools
will increase, if anything, in times of ubiquitous, pervasive, wearable and implantable
technologies. While novel technologies can affect our capabilities and our view of
ourselves as individuals and as a species, I believe that some aspects of humanity will
remain constant. Chief among them is the certainty that we will remain social beings,
conscious of the possibility and reality of suffering, but also endowed with plans and
hopes for a good life. We strive for happiness and seek to flourish in the knowledge
that we will always be negotiating the question: how exactly can flourishing best
be achieved? Technology can promote as well as reduce our flourishing. Our task
is therefore to ask how novel technologies can affect flourishing and what we can
do individually and collectively to steer such technologies in directions that support
flourishing. I hope that this book will help us make positive use of AI and move
towards a good, technology-enabled world.

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the Creative Commons license and
indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.