Machine Morality
121
This robustness/flexibility trade-off [7] points out one important observation:
natural selection might not have selected for a general-purpose control system,
but for a combination of several specialised systems that cooperating together
might overcome their intrinsic individual limitations.
4 Aversive and Appetitive Foundations of Moral
Decision-Making
Recent studies diving into the computational and neural substrates of moral
decision-making experience the influence of neuroeconomics in their experimental design and therefore, in the models that are developed. For instance, in a
series of experimental setups, a group of researchers has tried to investigate
assignment of value to certain properties of a moral decision, such as inflicting
harm to oneself or to others, through offering economic rewards in exchange for
varying levels of electrical shocks [11,47]. Behavioral and neuroimaging results
showed that subjects were preferring to harm themselves over others for profit
and this moral preference was associated with diminished neural responses in
value-sensitive regions to profit accrued from harming others [11]. Computational
utilitarian models developed to explain these results build upon the notion of
harm aversion. According to this view, harm aversion bias represents our intrinsic
avoidance of harming others and affects the way the internal utility computations
are integrated into moral decision-making. The notion of harm aversion can also
help to explain the puzzling results of the push–trapdoor divergence mentioned
before, and it might be an important ingredient of our moral cognition.
However, a computational model and a theoretical view on morality built
upon a negative proposition (harm avoidance/inhibitory behavior) would hardly
provide a comprehensive picture. In the end, if some part of our moral behavior
can indeed be based on a Pavlovian aversive reaction mechanism, other aspects
of morality could as well be grounded in an intrinsic preference for appetitive
stimuli. Such hypothesis would be backed up by substantial evidence obtained
from one-shot prisoner’s dilemma experiments where human subjects tend to
strive for cooperation as a default intuitive option [9,32–34].
In fact, anthropological accounts of the origins of morality do indeed provide
us with a view on morality that stems from a positive proposition: cooperation.
Cooperation in large groups of unrelated individuals is believed to be an evolutionary developed human feature [37]. In other words, it is evolutionary beneficial
for humans to genuinely prefer moral outcomes because it facilitates cooperation
[46]. Among recent theoretical proposals attempting to explain morality, one in
particular might seem promising as it integrates conceptual diversity of previous literature into one single theoretical framework capable of making testable
predictions: Morality-as-Cooperation (MAC) [12]. MAC defines morality as a
collection of biological and cultural solutions to the problems of cooperation and
conflict recurrent in human social life. Grounded in the mathematical theory
of cooperation (the theory of nonzero-sum games), MAC creates a principled
121
This robustness/flexibility trade-off [7] points out one important observation:
natural selection might not have selected for a general-purpose control system,
but for a combination of several specialised systems that cooperating together
might overcome their intrinsic individual limitations.
4 Aversive and Appetitive Foundations of Moral
Decision-Making
Recent studies diving into the computational and neural substrates of moral
decision-making experience the influence of neuroeconomics in their experimental design and therefore, in the models that are developed. For instance, in a
series of experimental setups, a group of researchers has tried to investigate
assignment of value to certain properties of a moral decision, such as inflicting
harm to oneself or to others, through offering economic rewards in exchange for
varying levels of electrical shocks [11,47]. Behavioral and neuroimaging results
showed that subjects were preferring to harm themselves over others for profit
and this moral preference was associated with diminished neural responses in
value-sensitive regions to profit accrued from harming others [11]. Computational
utilitarian models developed to explain these results build upon the notion of
harm aversion. According to this view, harm aversion bias represents our intrinsic
avoidance of harming others and affects the way the internal utility computations
are integrated into moral decision-making. The notion of harm aversion can also
help to explain the puzzling results of the push–trapdoor divergence mentioned
before, and it might be an important ingredient of our moral cognition.
However, a computational model and a theoretical view on morality built
upon a negative proposition (harm avoidance/inhibitory behavior) would hardly
provide a comprehensive picture. In the end, if some part of our moral behavior
can indeed be based on a Pavlovian aversive reaction mechanism, other aspects
of morality could as well be grounded in an intrinsic preference for appetitive
stimuli. Such hypothesis would be backed up by substantial evidence obtained
from one-shot prisoner’s dilemma experiments where human subjects tend to
strive for cooperation as a default intuitive option [9,32–34].
In fact, anthropological accounts of the origins of morality do indeed provide
us with a view on morality that stems from a positive proposition: cooperation.
Cooperation in large groups of unrelated individuals is believed to be an evolutionary developed human feature [37]. In other words, it is evolutionary beneficial
for humans to genuinely prefer moral outcomes because it facilitates cooperation
[46]. Among recent theoretical proposals attempting to explain morality, one in
particular might seem promising as it integrates conceptual diversity of previous literature into one single theoretical framework capable of making testable
predictions: Morality-as-Cooperation (MAC) [12]. MAC defines morality as a
collection of biological and cultural solutions to the problems of cooperation and
conflict recurrent in human social life. Grounded in the mathematical theory
of cooperation (the theory of nonzero-sum games), MAC creates a principled