-
The last several years have seen an intensification of discussions on the ethical aspects
of algorithms and artificial intelligence (AI) more generally. The heightened focus on
the ‘AI Ethics’ is a consequence of the fact that AI has started to appear in public
discourse as a powerful force shaping people’s lives. The term ‘AI’ is used in the debate
loosely, to refer to algorithmic technologies that rely on data and encoded assumptions
to identify patterns in a given domain or a topic area, in order to make it
comprehensible. Recently, various organizations and initiatives have produced a torrent of ethical
statements. Among these, the core principles of bioethics (beneficence,
non-maleficence, autonomy and justice) remain prominent. In addition, the principle of
explicability, incorporating intelligibility and accountability, is introduced as a response to the
overarching concern of the traceability of algorithmic operations
(Floridi et al. 2018;
Mittelstadt et al. 2016)
. At the same time, calls for moving ‘from principles to practices’
are becoming more frequent. Such a step entails the development of methods and tools
for the ‘applied AI ethics’. These are supposed to prompt engineers of algorithmic
systems to reflect on the impacts of their solutions on the ‘end users’, and on the ways in
which those impacts could be mitigated by certain design decisions at different stages
of development. The success of such tools is, nevertheless, premised on the increased
coordination between various stakeholders, coming both from within and outside of
developer communities
(Morley et al. 2019)
.
Unsurprisingly, the ‘AI Ethics’ debate has been critiqued extensively by social
scientists. The approach is said to reduce ethics to a technical or design issue while keeping
the status quo of current business practices (Greene, Hoffman and Stark 2018). The
lack of legal regulation has generated talk of ‘ethics washing’, referring to the fact that
while companies exercise self-regulation, they can define how they implement ethical
principles
(Wagner 2018)
. This has led to a call for abandoning the discussion on AI
ethics altogether and instead concentrating on questions of social inequalities and social
justice
(Sloane 2019)
.
Calls for disrupting a debate by ending it can be seen as a form of ‘fearless speech’:
no attempts for further dialogue are left open
(Englund 2018)
. To engage with the ethics
debate in a more productive manner, we propose an approach to algorithmic systems
and AI as ‘matters of care’
(Puig de la Bellacasa 2011; 2017)
. Care involves ‘everything
we do to maintain, continue, and repair ‘the world’ so that we can live in it as well as
possible’
(Tronto 1993, p. 103)
. This means that care includes all those (devalued)
‘productive doings that support liveable relationalities’
(Puig de la Bellacasa 2011, p. 93)
.
Caring entails not only ‘assembling concerns’
(Latour 2005)
, but also addressing those
not assembled, as well as adding new layers of concerns
(Puig de la Bellacasa 2011;
2017)
. Thus, care, focusing on doings and relationalities, is situated at the core of the
social scientific interest in exclusion and power dynamics.
To demonstrate what the move towards care and caring in the AI ethics debate might
mean, we discuss one possible way to promote it. After completing the
AlgorithmWatch report, Finland section
(Ruckenstein and Velkova 2019)
, we used the
empirical cases described in the report as a conversation opener in three workshops. With
this move, we acknowledged the openness of ethical and political questions concerning
algorithmic systems and opened a space for critical reflection. We learned that the
negative reactions concerning automated decision making were mainly focused on citizen
monitoring and its imagined consequences. For instance, the piloting of the uses of data
analytics in the public sector, with the aim of predicting future child service needs,
provoked fears about a deepening surveillance society and algorithmic control. Thus,
with the move to care, we can intensify the awareness of how technologies shape the
imaginaries of everyday lives. It prompts us to acknowledge that we live
interdependently with algorithmic systems rather than separated from them. In doing so, we
can expand the discussion from the focus on ethical algorithms to ethical living with
algorithms. Our workshop participants did not care about the specifics of algorithms,
but they cared deeply about the consequences of AI for their everyday lives and future
interactions as citizens.
The lens offered by care is, hence, suited for exploring human-technology relations,
but also for caring for them. In line with this, attention to care expands the debate on
AI ethics to issues currently excluded and neglected, involving maintaining and
repairing our shared life with algorithms. A credit scoring controversy, discussed in the
AlgorithmWatch report, underlines both citizens’ and ombudsman’s role in resourcefully
using existing legal and political tools to combat harms generated by algorithmic
systems. This brings to the fore that harms caused by algorithms can and need to be
repaired. Care introduces into the ethics discussion the erased human involvements with
technologies, from production, implementation and maintenance, to repair
(Suchman
2007)
. Care is a starting point for exploring the complexities of AI ethics by
re-establishing the human as a critical and creative actor, deeply interwoven with current
technological arrangements. If we aim for an ethical life with algorithms, we need to know
the consequences of AI systems intimately, as well as work with those systems, and
shape them.