Artificial Life and Robotics (2019) 24:291–296

https://doi.org/10.1007/s10015-019-00525-1

ORIGINAL ARTICLE

Artificial intelligence, ethics and human values: the cases of military

drones and companion robots
Thibault de Swarte1 · Omar Boufous1 · Paul Escalle1

Received: 26 March 2018 / Accepted: 11 December 2018 / Published online: 29 January 2019
© International Society of Artificial Life and Robotics (ISAROB) 2019

Abstract
Can artificial intelligence (AI) be more ethical than human intelligence? Can it respect human values better than a human?
This article examines some issues raised by the AI with respect to ethics. The utilitarian approach can be a solution, especially
the one that uses agent-based theory. We have chosen two extreme cases: combat drones, vectors of death, and life supporting
companion robots. The ethics of AI and unmanned aerial vehicles (UAV) must be studied on the basis of military ethics and
human values when fighting. Despite the fact that they are not programmed to hurt humans or harm their dignity, companion
robots can potentially endanger their social, moral as well as their physical integrity. An important ethical condition is that
companion robots help the nursing staff to take better care of patients while not replacing them.

Keywords Ethics · Artificial intelligence · Human values · Companion robots · Military drones · UAV

1 Introduction embedded in AI (Part 4). Finally, we will look at the cases

Can artificial intelligence (AI) be more ethical than human
intelligence? Can it respect human values better than a
human? Can a utilitarian approach using artificial agents
promote ethics? Even if this article is an exploratory research
on the above issues, we try to answer two difficult questions
raised by AI with respect to ethics. That is why we have
chosen two very extreme cases: combat drones, vectors of
killing, and companion robots, life supporting.
Contrary to a pilot fighter, a military drone is not subject
to stress during combat but would it be possible for it to
respect human values? Can companion robots help improve
the well-being of an increasing number of older people?
We will present the context of the Ethicaa team funded
by the Agence Nationale de la Recherche (ANR/National
Research Agency, France (ANR-13-CORD-0006) in part 2.
We will then discuss the need for an ethics of AI (Part 3).
Then, we will study the possibility of ethics and morality
This work was presented in part at the 23rd International
Symposium on Artificial Life and Robotics, Beppu, Oita, January
18–20, 2018.
* Thibault de Swarte
Thibault.deSwarte@imt‑atlantique.fr
1 on the context in which a drone or a companion is used.
IMT Atlantique, LASCO Laboratory, Rennes, France
of unmanned aerial vehicles (UAV) in part 5 and companion
robots (Part 6).
The key point of this article is that relations between AI
and humans must be discussed from an ethical and human
values perspective based on cases examined in their specific
contexts rather than too general perspective. This is the main
originality of the article below, based on the cases of mili-
tary drones and companion robots.
2 The context: Ethicaa’s research
programme
Machines and agents have an increasing number of
autonomous functions and consequently are less and less
supervised by human operators or users. Therefore, when
machines interact with humans, we need to ensure that they
do not harm us/them or threaten our/their autonomy, espe-
cially decision autonomy. Consequently, the question of an
ethical regulation or control of such autonomous agents is
raised. This has been discussed by several authors including
Wallach and Allen. As stated by Picard, the greater the free-
dom of a machine, the more it will need moral standards [1].
In this article, we have chosen to focus on “context ethics”
rather than broad moral standards. Each ethical rule depends

13
Vol.:(0123456789)
292
The objectives of the Ethicaa project are twofold: (1)
defining what should be a moral autonomous agent and a
system of moral autonomous agents, and (2) defining and
resolving the ethical conflicts that could occur. Ethical con-
flicts are characterized by the absence of an optimal solu-
tion: there may only be outcomes more desirable than others.
Nevertheless, when a decision must be made it should be an
informed decision based on an assessment of the arguments
and values at stake.
3 The need for ethical AI
3.1 AI definition
Artificial intelligence can be described by a universal triad,
namely the data brought by the environment, the opera-
tions defined as the logic which mimic human behavior and
finally, a control phase aiming at retroacting over its previ-
ous actions. Its definition is essentially based on two com-
plementary views: one focused on the behavior and how it
acts, especially as a human and the other which emphasizes
the reasoning processes and how it reproduces human skills
[2, 3]. However, both points of view insist on the rational
behavior that an AI must have. Moreover, it is important
to pay attention to which kind of AI we are dealing with:
strong or weak AI [4]. Weak AI, also known as narrow AI
is shaped by behaviors answering to observable and specific
tasks that may be represented by a decisional tree. On the
other side, the strong AI, or artificial general intelligence,
can copy human-like mental states. For this type of AI, this
means that decision abilities or ethical behavior are issues
that need to be taken care of. Finally a strong AI could find
the closest solution to the given objective and learn with an
external reversion.
The latter is the one that is posing unprecedented prob-
lems that researchers are just starting to study. In fact, a
system embedding strong AI is able to learn without human
assistance or injection of additional data since the AI algo-
rithm generates its own knowledge. Therefore, the exterior
observer or the user of such agents will no longer know what
the AI knows, what it is capable of doing nor the decisions
it is going to take. Hence the need to establish an ethical
framework that defines an area of action and prevents the
system from taking decisions contrary to the ethics.
3.2 How to implement ethics rules?
In order to implement ethical rules, there are two approaches.
The first one named top down approach is based on ethi-
cal rule-abiding machines [5, 6]. The strategy is to respect
unconditionally the ethical principles related to morality
such as “Do not kill”. However, without understanding the
13
Artificial Life and Robotics (2019) 24:291–296
potential consequences of the empirical decisions taken,
an AI system creates numerous approximations that are a
significant drawback. This can make rules conflict even for
the 3 laws of robotics [7]; it may also lead to unintended
consequences due to added rules [8]. It should be noted that
even inaction can be taken into account for injuring humans.
Moreover, the complexity of interaction between humans’
priorities may lead to interpersonal inappropriate compari-
sons of various added laws [9]. The second method called
bottom up, focuses on case studies in order to learn gen-
eral concepts. The case studies make it possible for a strong
AI to autonomously learn wrong and biased principles and
even generalize them by applying them in new situations it
encounters. These types of approaches are considered to be
dangerous. In fact, in this process of learning, basic ethical
concepts are acquired through a comprehensive assessment
of the environment and its compliance with previous knowl-
edge. This is done without any bottom-up procedure. The
result of this learning will be taken into account for future
decision-making [10, 11]. Eventually, in the case where an
AI algorithm is facing a new situation it has not encountered
before, the extrapolation without a control phase may result
in perilous situations for humans [12].
4 AI embedding moral and ethical
principles
A utilitarian approach of ethics consists in choosing in the
case of a set of possibilities the solution that leads to the
action consequently maximizing intrinsic good or net pleas-
ure [13]. This involves quantifying God or Evil from a given
situation. However, certain situations supported by ethical
reasons with an empirical study may prohibit the combined
execution of certain actions. These complex cases are at the
origin of dilemmas and the ethical principles do not make
it possible to establish a preference. Therefore, autonomous
agents need to be endowed with the ability to distinguish the
most desirable option in the light of the ethical principles
involved.
To achieve this goal, this article proposes in the following
subsections a method called a utility function as a mean of
avoiding ethical dilemmas.
4.1 Using a utility function to help agents make
ethical decisions
In order to achieve this goal, a number of solutions have
been proposed [14]. One of them it the utility function,
also known as objective function. This function is used to
assign values to outcomes or decisions. The optimal solu-
tion is the one that maximizes the utility function. This
approach based on quantitative ethics determines which
Artificial Life and Robotics (2019) 24:291–296
action maximizes benefit and minimizes harm. Its objec-
tive is to make it possible for an AI algorithm to take the
right decisions particularly when it encounters an ethical
dilemma.
From a mathematical point of view, the utility function
takes a state or a situation as an input parameter and gives
as a result an output which is a number [15]. This number
is an indication of how good the given state or situation
is for the agent. The agent should then make the decision
that leads to the state that maximizes the utility function.
For instance, let us take the case of an autonomous
vehicle, and let us assume that the car is in a situation
where harm is unavoidable, and that it would inevitably
hit either two men on the road or crash into a wall killing
the passenger it is carrying [16]. Based on our previous
utilitarian ethics definition, the decision that will minimize
harm is the one that will lead to kill as few people as possi-
ble. Therefore, the car should crash and kill the passenger
to save the two pedestrians because the utility function of
this outcome is the highest. The same reasoning applies
for military drones when they have to choose between mul-
tiple outcomes that involve moral and ethical principles.
Autonomous cars embedding AI algorithms using the
utility function are not yet marketed. Some models avail-
able to the general public have an autopilot mode that still
requires the presence of a human being behind the steer-
ing wheel who will make a decision in case of a problem.
Fully autonomous cars still ride in test environments [17].
In the near future, the people who are likely to buy this
type of car will primarily be public institutions such as
municipalities. For instance, the city of Helsinki is test-
ing an autonomous bus line, RoboBusLine which carries
passengers on a defined road with a limited speed and an
autonomous shuttle is also in service in Las Vegas [18].
However, these are still prototypes in a test phase with
an operator on board. The other customers that may be
interested in using autonomous vehicles are companies
that make deliveries given the advantage of automating
the tasks resulting in cost reduction and efficiency. In fact,
Amazon, FedEx and UPS are investigating solutions for
driverless trucks.
The utility function is currently under investigation as
an active solution to avoid ethical dilemmas non-modifying
on-policy [19]. Autonomous robots are expanding and the
aim is not only to deal with ethical dilemmas but also to
reduce uncertainty by quantifying problems such as explora-
tion or unknown mapping; both can be stochastically defined
(Shannon or Rényi’s entropy) [20, 21]. Describing and tak-
ing actions in a world incompletely defined can be done
with the help of estimators but utility functions describe the
perceptual state in line with the rules, and an active strategy
can hence be implemented. This is already done for robot
vision, for example [22].
293
4.2 Limits and dangers of the utilitarian approach
The approach previously described and consisting in quan-
tifying situations and assessing them with a utility function
through a model has its own limits as far as strong AI is
concerned.
For weak AI, engineers at the design stage can imple-
ment decision trees to establish rules. They can anticipate
the behavior of the AI more easily. On the other hand, as
mentioned in Sect. 2, advanced AI systems learn directly
from the environment and adapt accordingly. By doing so,
an external observer cannot always predict or anticipate the
actions of such systems [23]. This is true for the Alpha Go
algorithm that is taking decisions and implements strategies
even experts in the game cannot understand although it leads
to an optimum solution. The intelligent agent behaves like
a black box whose internal functioning is unknown. This is
particularly dangerous when it comes to autonomous vehi-
cles or UAV drones that put human life at stake. Using only
a utility function to decide whether or not a UAV could be
used in an armed conflict could be considered as a war crime
(see below part 5, Sect. 2).
Indeed, it is essential to test AI algorithms in different
environments [23] and cover as many situations as possible
before they are registered for use. This involves confront-
ing algorithms with different situations and ensuring they
behave properly by taking the most ethical decisions possi-
ble. It will then be possible to identify anomalies and correct
them immediately.
5 Artificial intelligence, ethics and human
values: the case of UAV
The ethics of artificial intelligence are post-Kantian [24] in
the sense that they do not accept any opposition between nat-
ural and cultural sciences. They actually seek to overcome
such an opposition. In the “Ethicaa” context, it is hoped that
a contextual evaluation will be made for each particular situ-
ation. A definition of a legal framework and the enactment
of UAVs matter.
Ethical macro-dilemmas are linked to the fact that a
democracy cannot, because of its human values, use drones
without precise rules of engagement. Micro-dilemmas are
the operational processes by which a drone performs an
action. Do cameras, for example, provide a clear view of
the target? If not, what should the operator do? The meso-
dilemmas are the most interesting to study. They are at the
articulation of the two types of dilemmas above. Georges
Lucas, Professor of Ethics at the Naval Postgraduate School
has expressed caution primarily about marrying “strong
Artificial Intelligence with full lethality (...). I do not want
13
294
robots making lethal targeting decisions entirely on their
own” [25].
In general, the military doctrine emphasizes that ethi-
cal choices are ultimately a matter of values, organized by
the military rules and their hierarchy. Weber [26], with his
vision of the “ethics of responsibility” and his model of
“legal rational decision” making seems best suited to deal
with ethical conflicts. It would indeed be irrelevant to con-
sider an artificial agent building an ethical decision in the
place of a human agent, because ethics are precisely what are
produced by natural human intelligence. If we seek to adapt
Weber’s [27] framework to that of Ethicaa, we should also
ask ourselves the question of “scientific ethics and the spirit
of artificial intelligence”: scientific ethics has therefore to be
as independent as possible from political or religious ethics.
Our colleague Ganascia [28] defines the “spirit” of arti-
ficial intelligence as follows: “it shows how this epistemo-
logical view opens on the many contemporary applications
of artificial intelligence that have already transformed-and
will continue to transform-all our cultural activities and our
world.
5.1 Traditional military ethics and combat drones
Danet et al. [25] believe that the principles of military ethics
are generally stable over time since the “War Art” (Machi-
avelli 1520) [29] and that a specifically human honor code is
required. For Gérard de Boisboissel interviewed by Thibault
de Swarte, since the age of Chivalry in Europe, the com-
mand chain is still ultimately responsible and no artificial
intelligence can replace it. The ethical dilemma is always
that of the high-ranking officer who must take responsibil-
ity for ordering a drone to kill in accordance with “military
grandeur and servitude” [30]. We consider Mishima and his
ethic of the Samurai (“Hagakuré”) [31] dating back to the
eighteenth century to be part of the same military values
tradition.
5.2 UAV and military ethics in the twenty‑first
century
An armed drone does not create a break from other weapons.
On the contrary, there is a reversibility that the previous
weapon systems did not have (artillery, fighter jets, etc.).
For the French military expert Pierre Servent interviewed by
Thibault de Swarte, the question of whether a combat drone
can be ethical or not can be ironical and serious at the same
time. It should in fact be asked for all types of weapons. The
key question is, “is the shooter under a democratic control?”
A democracy cannot delegate the right to kill to a machine.
The French position is characterized by precaution. This
country wants to develop a drone use doctrine as a prelimi-
nary step. In contrast, the USA have recently trained more
13
Artificial Life and Robotics (2019) 24:291–296
UAV pilots than jet fighter pilots. They have afterwards
developed a doctrine. Democratic values have sometimes
been set aside in the name of short-term efficiency. Later,
in the 2010s, ethical questions were asked by the press and
public opinion. Is it acceptable to consider that an artificial
agent could take an ethical decision in the place of a human
agent, while ethics are precisely what is produced by human
intelligence?
Finally, a drone is not in essence “more” or “less” ethical.
The military process is often the same whether it is a drone
or a combat jet. An important problem is the reliability of
the equipment and the reversibility capability it offers or
not. In any case, avoiding the “Fire and Forget” is necessary.
Combat aircraft may not hit.
AI can be a condition for better ethics and respect of
human values, because a UAV has less stress than a human
pilot in a certain number of war operations.
6 The case of companion robots
6.1 The global context
The dignity of robots and the human person is sometimes
the topic of intense academic debates. However, it had never
been addressed by the case-law of the European Court of
Human Rights before 2014 [32]. It is therefore an emerging
issue for which research is fully legitimate in order to be able
to inform the legislator in due course.
Europe in general and France in particular show certain
skepticism towards companion robots. This skepticism
is taken into account by the European Parliament. At the
beginning of 2017 its Committee on Legal Affairs adopted a
text calling on the executive body—the European Commis-
sion—to take some measures to control robots and artificial
intelligence and to settle questions regarding their compat-
ibility with ethical standards and reliability. If UAVs are
not concerned by this text proposal, companion robots are.
Companion robots have been introduced in response to
problems emerging from the ageing of the population and
to address therapeutic needs. Some people need care and
others companionship. Today, companion robots assist autis-
tic children or the elderly, for example those with dementia
who need companionship in retirement homes. The lack of
medical personnel in health services sometimes explains the
use of such robots. In retirement homes in Denmark, a thera-
peutic robot called Paro plays the role of a pet. Its sensors
allow it to respond to the cuddles of patients, by moving its
tail, body and eyes. This robot mimics the behavior of a seal.
And as with all “carebots”, it operates in the intimacy of
the people who use it on a daily basis. Although it has been
shown that this type of robot relieves older people on a psy-
chological point of view [33], its proximity is problematic
Artificial Life and Robotics (2019) 24:291–296
on two levels: on one hand, it can jeopardize the privacy
and intimacy of the person. On the other hand, it leads to a
reduction in human contact, and potentially to social isola-
tion in the longer term. Both aspects are discussed in the
two following parts.
6.2 The issue of human dignity and privacy
The right to privacy is a fundamental right guaranteed by
the Article 8 of the European Convention on Human Rights
and by the Universal Declaration of Human Rights. If robots
are used to lift—as does the nursing robot Robear [34]—and
move people in a way that implies that they are objects,
it could have the effect of reducing their self-esteem and
making them feel humiliated. By doing so, this harms their
dignity [35].
The real problem of robots is the program they contain.
This program is written by a computer scientist with his
own bias, his own vision of the world and especially his
imperfections and flaws. And the question is not whether
the algorithms are ethical or not but whether the way they
are used is good or bad; if they fulfill the human ethical
intentions and values as humanly and socially as possible.
6.3 Ethics, human values and companion robots
There are currently no procedures of control or these algo-
rithms that are trusted by users when new forms of discrimi-
nation, censorship, impasses, errors, artifacts, social norms,
false information and predictions appear. We are unable to
manage these problems because the legal system is not yet
adapted to the exponential AI development and its daily
applications. On the contrary, a jurist such as Alain Ben-
soussan heard at the EOGN research center in Paris sup-
ports the fact that we should protect robots from humans’
interventions on behalf of robots dignity.
The vast majority of companion robots, such as NAO are
connected to the Internet. Today, data are recorded by inte-
grated cameras and microphones and sent to remote servers.
Often, the owners of these robots tend to forget why their
personal information is recorded and stored in the cloud. The
current programming mode is the one that tends to make
robots indispensable in everyday activities. But the robot
must not become a “prescriber” of behavior according to
Serge Tisseron. This would mean that the human would be
depending on the robot and that the robot would reify the
human, which would be contrary to human dignity. In the
case of “carebots” used in hospitals, it is necessary that they
preserve and guarantee the autonomy of the patient with
regard to the medical staff and treatments. This requires
the consent of the patient or his family and the possibil-
ity for him to retract and stop his treatment without con-
flict. Detailed information should also be given specifying
295
the circumstances of robot use for the patient. Moreover,
every computing device that is created is hackable, includ-
ing neural implants and military systems. This means that
robots have security flaws too. For instance, in 2016, Chi-
nese hackers managed to remotely control an autonomous
car. By doing so, they had full control of all commands of
the vehicle and they were able to deliberately cause an acci-
dent. One can imagine how the consequences of such actions
could be lethal. Therefore, even if they are intelligent and
sophisticated, robots can be hijacked and their source code
can be modified to be used for criminal purposes. Despite
that they are not programmed to hurt humans or harm their
dignity, robots can potentially endanger their social, moral
and physical integrity.
7 Conclusion
This article has examined some issues raised by artificial
intelligence with respect to ethics and human values. This
is an exploratory research on the above issues.
The utilitarian approach is a solution, especially one that
uses agent-based theory. Using a utility function can help
agents make ethical decisions.
We have here chosen two extreme and very different
cases: combat drones, vectors of death, and companion
robots, in favor of life.
The ethics of AI must be studied in all circumstances
on the basis of human values when fighting and in military
ethics. Combat UAVs are part of the contemporary history
of such ethics that research must continue to develop and
universalize, thanks to concepts like shot reversibility.
Despite the fact that they are not programmed to hurt
humans or harm their dignity, companion robots can poten-
tially endanger their social, psychological as well as their
physical integrity. A general ethical condition, in line with
human values, is that companion robots can help the nursing
staff to take better care of patients but do not replace them.
We hope that our research will continue on these topics
and other emerging ones in our societies like ethics applied
to autonomous cars, digital platforms or internet of things
(IOT).
References
1. Picard RW (2003) Affective computing: challenges. Int J Hum
Comput Stud 59(1):55–64. https​://www.scien​cedir​ect.com/scien​
ce/artic​le/abs/pii/S1071​58190​30005​21. Accessed 27 Jan 2019
2. Nilsson NJ (1980) Principles of artificial intelligence,
pp 17–18. https​://www.sprin​ger.com/la/book/97835​40113​409.
Accessed 27 Jan 2019
3. Russell S, Norvig P (1995) Artificial intelligence: a modern
approach, pp 4–5. https​://ready​forai​.com/downl​oad/artif​icial​
13
296
-intel​ligen​ce-a-moder ​n-appro​ach-3rd-editi​on-pdf/. Accessed 27
Jan 2019
4. Bringsjord S, Schimanski B (2003) What is artificial intelligence?
Psychometric AI as an answer, p 6. https​://www.ijcai​.org/Proce​
eding​s/03/Paper​s/128.pdf. Accessed 27 Jan 2019
5. Powers TM (2006) Prospects for a Kantian machine,
pp 48–50. https:​ //www.academ ​ ia.edu/314677​ 71/Prospe​ cts_for_a_
Kanti​an_Machi​ne. Accessed 27 Jan 2019
6. Hanson R (2009) Prefer law to values. http://www.overc​oming​
bias.com/2009/10/prefe​r-law-to-value​s.html. Accessed 27 Jan
2019
7. Asimov I (1950) “Runaround”. I, Robot, p 40. https:​ //www.ttu.ee/
publi​c/m/mart.../Isaac​_Asimo​v_-_I_Robot​.pdf. Accessed 27 Jan
2019
8. Pettit P (2003) Akrasia, collective and individual. https​://core.
ac.uk/downl​oad/pdf/15661​6471.pdf. Accessed 27 Jan 2019
9. Wallach W, Allen C, Smit I (2008) Machine morality: bottom-
up and top down approaches for modelling human moral facul-
ties, pp 570–579. https​://link.sprin​ger.com/artic​le/10.1007/s0014​
6-007-0099-0. Accessed 27 Jan 2019
10. McLaren B (2006) Computational models of ethical reasoning,
pp 30–32. https​://www.cs.cmu.edu/~bmcla​ren/pubs/McLar​en-
CompM​odels​Ethic​alRea​sonin​g-MachE​thics​2011.pdf. Accessed
27 Jan 2019
11. Guarini M (2006) Particularism and the classification of moral
cases, pp 23–26. https​://ieeex​plore​.ieee.org/docum​ent/16679​49/.
Accessed 27 Jan 2019
12. Muehlhauser L, Helm L (2012) Intelligence explosion and
machine ethics. https:​ //intell​ igenc​ e.org/files/​ IE-ME.pdf. Accessed
27 Jan 2019
13. Sinnott-Armstrong W (2015) Consequentialism, Stanford Ency-
clopedia of Philosophy. https:​ //stanfo​ rd.librar​ y.sydney​ .edu.au/entri​
es/conse​quent​ialis​m/. Accessed 27 Jan 2019
14. Anderson M, Leigh Anderson S (2011) Machine ethics. https​://
www.cambr​idge.org/core/books​/machi​ne-ethic​s/D7992​C92BD​
465B5​4CA0D​91871​398AE​5A. Accessed 27 January 2019
15. Hibbard B (2011) Model-based utility functions. https:​ //arxiv.​ org/
vc/arxiv​/paper​s/1111/1111.3934v​1.pdf. Accessed 27 Jan 2019
16. Bonnefon J-F, Rahwan I, Shariff A (2015) Autonomous vehicles
need experimental ethics: are we ready for utilitarian cars?. https​
://www.resea​rchga​te.net/publi​catio​n/28284​3902_Auton​omous​
_Vehicl​ es_Need_Experi​ menta​ l_Ethics​ _Are_We_Ready_​ for_Utili​
taria​n_Cars. Accessed 27 Jan 2019
17. Tian Y, Pei K, Jana S, Ray B (2018) DeepTest: automated testing
of deep-neural-network-driven autonomous cars. https:​ //arxiv.​ org/
abs/1708.08559​. Accessed 27 Jan 2019
18. Kirk B (2016) Business opportunities in automated vehicles. In
Journal of Unmanned Systems. http://www.nrcres​ earch​ press​ .com/
doi/full/10.1139/juvs-2015-0038#.XE3kH​88zYX​p
19. Everitt T, Filan D, Daswani M, Hutter M (2016) Self-modification
of policy and utility function in rational agents. https​://arxiv​.org/
abs/1605.03142​. Accessed 27 Jan 2019
20. Carrillo H, Dames P, Kumar V, José A. Castellanos (2017)
Autonomous robotic exploration using a utility function based
on Rényi’s general theory of entropy. https​://link.sprin​ger.com/
artic​le/10.1007/s1051​4-017-9662-9. Accessed 27 Jan 2019
21. Keren S (2017) Redesigning stochastic environments for maxi-
mized utility. https​://www.aaai.org/ocs/index​.php/AAAI/AAAI1​
7/paper​/downl​oad/14549​/14200​. Accessed 27 Jan 2019
13
Artificial Life and Robotics (2019) 24:291–296
22. Arindam Bhakta C, Hollitt WN, Browne M, Frean (2018) Utility
function generated saccade strategies for robot active vision: a
probabilistic approach. https​://link.sprin​ger.com/artic​le/10.1007/
s1051​4-018-9752-3. Accessed 27 Jan 2019
23. Hibbard B (2015) Ethical artificial intelligence. https​://arxiv​.org/
abs/1411.1373. Accessed 27 Jan 2019
24. Kant E (1785) Fondements de la métaphysique des mœurs. Feed-
books. http://fr.feedb​o oks.com/book/114/fonde​m ents​- de-la-
m%25C3%25A9t​aphys​ique-des-moeur​s. Accessed 25 Jan 2019
25. Danet D, Doaré R, Hanon JP, de Boisboissel G (2014) Robots
on the battlefield: contemporary issues and implications for the
future. Combat Studies Institute Press, France, p 301. https:​ //apps.
dtic.mil/docs/citat​ions/ADA60​5889. Accessed 27 Jan 2019
26. Weber M (1917/1949) The meaning of ‘Ethical Neutrality’ in
sociology and economics. In: The methodology of the social sci-
ences. https​://www.taylo​r fran​cis.com/books​/97813​51505​574/
chapt​ers/10.4324%2F978​13151​24445​-1. Accessed 25 Jan 2019
27. Weber M (1920) The Protestant ethic and the " spirit” of capital-
ism and other writings. https​://is.muni.cz/el/1423/podzi​m2013​/
SOC57​1E/um/_Routl​edge_Class​ics___Max_Weber​-The_Prote​
stant​_Ethic​_and_the_Spiri​t_of_Capit ​alism​__Routl​edge_Class​
ics_-Routl​edge__2001_.pdf. Accessed 27 Jan 2019
28. Ganascia J-G (2010) Epistemology of AI revisited in the light of
the philosophy of information. Knowl Technol Policy 23(1–2):57–
73. https​://link.sprin​ger.com/artic​le/10.1007/s1213​0-010-9101-0.
Accessed 27 Jan 2019
29. Niquet V (1988) L’Art de la guerre de Sun Zi, traduction et édi-
tion critique, Éditions Economica. https:​ //www.econom ​ ica.fr/livre​
-l-art-de-la-guerre​ -sun-zi-niquet​ -valeri​ e,fr,4,978271​ 78589​ 69.cfm
30. Vigny de A (1835) Grandeur et servitude militaires. http://www.
bouqu​ineux​.com/?ebook​s=82&Vigny​. Accessed 25 Jan 2019
31. Mishima Y (1985) le Japon moderne et l’éthique samouraï. Gal-
limard, Paris. http://www.galli​mard.fr/Catal​ogue/GALLI​MARD/
Arcad​es/Le-Japon​-moder​ne-et-l-ethiq​ue-samou​rai
32. European Court of Human Rights (1999–2014) Reports of judg-
ments and decisions—cumulative index 1999–2014. https:​ //www.
echr.coe.int/Docume​ nts/Index_​ 1999-2014_ENG.pdf. Accessed 25
Jan 2019
33. MacDonald B, Kerse N, Broadbent R (2013) The psychoso-
cial effects of a companion robot: a randomized controlled trial
Hayley Robinson. https​: //www.ifa-fiv.org/wp-conte​n t/uploa​
ds/2015/11/2013-Paro-loneli​ ness-​ RCT.pdf. Accessed 27 Jan 2019
34. Mullan TMC (2016) The guardian—how a robot could be grand-
ma’s new career. https​://www.thegu​ardia​n.com/techn​ology​/2016/
nov/06/robot​-could​-be-grand​mas-new-care-assis​tant. Accessed
Oct 2016
35. Sharkey A, Sharkey N (2010) Granny and the robots: ethical
issues in robot care for the elderly. https​://link.sprin​ger.com/artic​
le/10.1007/s1067​6-010-9234-6. Accessed 27 Jan 2019
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.