Page 152 - Proceedings of the 2018 ITU Kaleidoscope

2018 ITU Kaleidoscope Academic Conference

High) in a design. The four different possibilities that may For a designer to lend a machine (through the algorithms
emerge for a designer of ML algorithm can be classified as designed by her) a low SQ, constant IQ but a higher level
- ‗Mechanical Learner‘, ‗Cognitive Learner‘, ‗Ethical of EQ would make the machine ‗behave‘ empathetic
Learner‘ and ‗Ideal Ethical Master‘ as described below. towards others and so the learning process in this case
would be inclusive of remembering ethical frameworks of
Possibility 1: Mechanical Learner (LOW EQ, LOW previous thinkers and concepts such as Utilitarianism and
SQ) then understanding their applications in certain contexts.
Such machines would therefore, learn by themselves which
The first stage in machine learning is when the machine would make them susceptible to sometimes making
becomes a mere mechanical learner because it relies on unethical choices as well. The drawback would be that the
algorithms which are fed into its system by a designer who ethical parameters are general in nature and just like
inherently has a low EQ and low SQ. This implies that the humans may or may not find it ethically correct to stay true
designer may not understand and respond to human to a particular philosophy similarly the machine too would
emotions in a sensitive way or simply lacks empathy. The take the ‗best guess prediction‘ which will be partially
ability to feel for others is present but lacks empathetic correct.
response. The machine devoid of any emotional
understanding will thus not think of the implications of its Possibility 3: Ethical Learner (LOW EQ, HIGH SQ)
actions but by default will follow the commands as stated
in its algorithm. With a low EQ imbued in the machine by its designer, the
machine will not stay empathetic towards humans but with
a high SQ it would have awareness of how the action of the
Low EQ does not mean presence of negative feelings rather self (machine) affects others around it. The Talisman of
it means a neutral response towards another being. A low Gandhi (1958) is the best example here as it talks about
SQ means the value clarification is not fully mature. Self how when one is faced by a dilemma, one should think of
and social awareness is undermined. the implications of one‘s actions on the weakest person one
can think of [40]. Such a machine, with Low EQ and High
SQ, would be an ethical learner as it would be awakened to
the needs of others and therefore, would think in terms of
value clarification of the deeds performed by the self and
their larger implications on those who may be at the
receiving end.

Presumably, an ethical learner type of ML algorithms
would be more acceptable as it will have the capacity to
look beyond ‗the self‘ and will have inclination to be of
service to others (humans) and hence would inadvertently
follow Asimov‘s Zero Law of Robotics too. The
personality type of the designer, belonging to this quadrant
will never design machine algorithms that are ‗biased‘.
Such a machine will not voluntarily participate in
discrimination against anyone based on previously
highlighted markers such as race, class, sex, age etc.
Possibility 4: Ideal Ethical Master (HIGH SQ, HIGH
Figure 2 - Possible ethical dimensions for a designer EQ)
of machine learning algorithms
A high SQ and a high EQ (coupled invariably with the
constant IQ) is the best case scenario which would be most
Possibility 2: Cognitive Learner (LOW SQ, HIGH EQ)
ethical in nature, lending it almost an idealist dimension, an
utopian existence. With such a designer, inking her
This type of learning is acquired through an active use of
emotions, thought processes and sensory perceptions. algorithms, the machine in this quadrant would have
Psychologist Benjamin Bloom‘s Taxonomy of learning invariably mastered ethical dilemmas and would function
domains was altered in mid nineties to the domains in the flawlessly without ever jeopardizing or compromising on
someone‘s safety or life. A concern for harmony on earth
order as: remembering, understanding, applying, analyzing,
evaluating and creating [17]. Humans have evolved to this and using ideas for careful utilization of resources makes
existing stage through cognitive learning approach and this this an ideal situation in machine learning, leading to a self-
is what actually separates humans from machines. regulated system, suitable for all. Recapitulating, what has
been previously stated regarding the subjective nature of

– 136 –