Proposed Way to Inculcate Morality in Artificial Intelligence

Richa Kapoor Mehra

O. P. Jindal Global University, Sonipat, India

Keywords: Artificial Intelligence, Ethical Theories, Utilitarianism, Categorical Imperative, Virtue Ethics.

Abstract: Recently one of the most trending, memorable and unique news hitting the social media that showcased one

of the best uses of technology is in the form of the clad draped in a traditional saree to accept loan documents

from bank branch. Robotic company ASIMOV sent its saree- clad advanced robot SAYABOT to receive loan

documents from the bank. One of the main reasons for the popularity of such humanoids is due to their

capability of performing task in almost all sectors such as education, healthcare, hospitality and banking. In

the 21st century, human dependency on robots and other gadgets has increased drastically, as a result giving

rise to plenty of challenges. With AI becoming smarter day by day in taking decisions and getting involved

in other activities results into various ethical concerns. Few questions that one can ponder upon- can these

digitally empowered Bots who behave like humans and dress like humans, act in socially and morally

responsible way? Can Artificial Intelligence be said to be moral agents? Can artificial intelligence be said to

have the ability to discern right from wrong? Who bears responsibility of wrong doings by Artificial

Intelligence? Can we hold artificial intelligence accountable for its own actions? Is there a way to equip

artificial intelligent as artificial moral agents? This is where role of ethics comes in. The increasing use of AI

in 21st century in more efficient and faster way poses various challenges on society and its beings. Now one

of the significant question is what should be done to ensure that AI outbreak gives rise to positive results

which benefit the society as a whole? This article is an attempt to answer the above-mentioned questions by

showcasing ethics as a means to tame and instruct artificial intelligence to ensure they act in a morally

responsible way. The aim of this article is to determine which ethical theory is best suitable theory that would

provide results in favor of humanity. The first section of this article deals with consequentialist theory of

utilitarianism. In the second section an attempt has been made to comprehend how compatible artificial

intelligence would be with Aristotle’s virtue theory. The third and last section showcase the compatibility of

artificial intelligence with deontology theory of Immanuel Kant.

1 INTRODUCTION

Since time immemorial, morality is considered to be

the very essence of human beings. Morality is the

fundamental human trait that makes us responsible

moral agents. Human dignity and moral responsibility

are the key essence of human traits. Humans are

moral agents; they have the ability to discern from

right and wrong actions and are considered

accountable for their actions. With the growing use

and demand of artificial intelligence, it is imperative

to check the actions of AI and ensure that their actions

are ethically governed. Before we move forward, it is

imperative to first know what is meant by artificial

intelligence.

AI can be understood as intelligence that is

artificially created and that is demonstrated by a

machine. From Alexa to touch screen dispensers and

other gadgets, we can see our dependence on many

such intelligent items. The term AI can be understood

as intelligent agent that is designed and created by

human beings; is so intelligent that it perceives its

environment and becomes smarter. The most

common existing form if artificial intelligence is

simple AI, these are machines that rely on the

preprogrammed data or set of rules hence may not be

very efficient decision makers. On the other hand

there is Machine Learning, which permits machines

to acquire from the surroundings without being

overtly automated thus equip themselves as better

decision makers. This type of technology is often

referred as complex artificial intelligence; they have

better decision-making capabilities as they get

involved in integrating and scrutinizing data relating

to a specific task. Since these complex AI takes

Mehra, R.

Proposed Way to Inculcate Morality in Artiﬁcial Intelligence.

DOI: 10.5220/0012615400003739

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Artiﬁcial Intelligence for Internet of Things: Accelerating Innovation in Industry and Consumer Electronics (AI4IoT 2023), pages 319-321

ISBN: 978-989-758-661-3

319

decision on the bases of pre given data, so to ensure

they act in morally appropriate way; I suggest to

equip them with certain ethical theories. Thus, to

ensure best results we have to feed ethical data to the

AI so that they act in ethically smarter and better way.

Now the question that can be posed- how one become

aware of ethical data? To unfold this question, we will

explore various ethical theories and will discover

which theory yield the best outcome for humanity?

Ethical theories provide us with the foundation for

decision making. Ethical theories help in

distinguishing between right and wrong actions.

These theories help us in getting the unified account

of our moral obligations. If AI programmers instruct

AI with the most suitable ethical theories, then such

theories will benefit AI in taking morally right

decisions. To determine which ethical theory would

yield the best result it is imperative to unpack few

most relevant ethical theories.

2 CONSEQUENTIALIST

THEORY

The judgement of action as good or bad is determined

on the basis of the outcome of the action. An action

giving rise to good consequences is considered as

good action, on the other hand if the consequences of

an action are bad; the action is considered as bad. The

morality of action in consequentialist theory is

determined on the basis of its outcome. This theory

judges an action to be good or bad on the basis of its

consequences. For instance, most of people would

believe that lying is bad; but for consequentialist, if a

lie saves someone’s life then it is not bad to lie. There

are primarily two types of consequentialists-

utilitarianists and hedonists. One of the famous forms

of consequentialism is utilitarianism, it aims to

perform those actions which may lead to maximum

benefits to maximum number of people. In other

words, an action that brings more benefit than harm is

considered as good action, and the action that

produces more harm than good is considered as bad

action and should be avoided. Two famous English

thinkers who have defended utilitarianism include

Jeremy Bentham and John Stuart Mill. Utilitarian’s

doctrine can be expressed in the form of a single

principle, the greatest happiness principle: the

rightness of an action is determined by its

contribution to the happiness of everyone affected by

it. On the other hand, hedonists believe that an action

is good if it results into pleasure and it’s bad if it gives

rise to pain. To achieve the maxim of maximum

happiness for maximum number of people, for

utilitarians it is imperative to increase number of good

things in the world and to decrease the number of bad

things.

3 VIRTUE THEORY

This theory has gained popularity in past 20 years,

virtue theory is contrary to consequentialist theory.

Aristotle, the famous ancient Greek philosopher

advocated this theory. The prime emphasis of this

theory is development of one’s character, it focused

on the development of human qualities. Prime

questions emphasized by this theory includes - “How

should a person live?” and “What is the good life?”

and “What are values?”

Aristotle believed that virtues can be imbibed

through habit, according to him by practicing virtues

repeatedly a person develops moral character.

Through continuous practice of being truthful, just,

generous, compassionate and so on, a person

develops a decent ethical character. This results into

making correct choices specially during ethical

dilemma situations.

Aristotle then observes that a thing is called good

when it performs its function well same is applicable

to human beings; for him when a person channelize

reason and take decision based on reasoned inquiry,

they are believed to perform well. For Aristotle, man’s

highest function is to reason well and a good man is one

who reason’s well. However, virtue theory may not be

said to be compatible with artificial intelligence since

having virtuous traits in artificial intelligence does not

ensure that the decision making will always be in

favor of other human beings. In the up-coming section

deontological theory has been deliberated to explore

the compatibility of this theory with the artificial

intelligence.

4 DEONTOLOGICAL THEORY

In contrast to consequentialist theory, deontology

theory was proposed by Immanuel Kant, it

contemplates actions to be ethically right or wrong in

and of themselves, irrespective of their outcome; they

use universal rules to distinguish right from wrong.

The ethical theory of categorical imperative is core

ethical theory of Kant. The term categorical

imperative, basically means “absolute command.”

According to Kant, there is only one categorical

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

320

imperative, which he presents in three diverse

principles:

1: The law of universalizability: this principle

states that an action is considered to be right if it can

be universalized. For instance, breaking promise or

telling a lie; before performing these actions, a person

has to see whether it can be universalized. Kant puts

it this way: “Act as though the maxim of your action

were by your will to become a universal law of nature.

2: The Principle of Ends states that: “So act as to

treat humanity, whether in your own person or in that

of any other, in every case as an end and never as

merely a means.” To put it differently, this law states

that we should treat human beings as end in

themselves and never use them as a means to fulfill

some other end., To put it differently, individuals

should give respect to each human being and should

not use them as a commodity to accomplish the

desired results. Kant states that since we humans have

rational capacity or ability to reason we can make use

to reason to perform right actions.

3: The Principle of Autonomy this principle states

that we are free rational beings and due to this pre-

given rationality, we are able to determine the

differences between right and wrong actions. To

differentiate between right and wrong actions, we do

not have dependent upon others. However, for Kant

the onus lies on humans to discover and identify

differences between morally right and wrong actions.

We have to use our ability to reason to assist us in

implementing the categorical imperative and make

our own decisions, rather than relying on someone

else to tell us what to do. Kant puts it this way: “So

act that your will can regard itself at the same time as

making universal law through its maxims.

Now putting these ethical theories in context of

artificial intelligence decision- making, the

differences in the theories would result in the

difference in decision making. For instance, a

consequentialist AI might think that killing one evil

human being would be a just act if it results into

saving thousands of other human beings. However

contrary to the normative approach, a virtue theory AI

emphasizes on development of moral traits of an

individual. AI following deontological theory unlike

the consequentialist theory would regard the act of

killing as a wrong act.

5 CONCLUSION

With the advancement of technology and digital

revolution in place, it is believed AI to have rational

thinking capacity if it involves in a pre-given design

of logical thinking basis which it vindicates and

perform actions. If applied to AI, deontology theory

is most appropriate theory because of its ordering

actions and categorizing it on the basis of rational

dimensions of the other moral agents. It uses

universal law to distinguish right from wrong. For

Kant, humans are distinctly superior to other beings

due to its rational capacities. This status would also

apply to artificial intelligence. In my opinion, out of

the above mentioned ethical theories, AI must be

automated with deontological. Since AI have the

potential to make incredibly complex moral

decisions, it is important that humans are able to

identify the logic used in a given decision in a

transparent way, so as to accurately determine the

morality of the action in question.

REFERENCES

Abelson, Raziel and Kai Nielsen. “Ethics, History of” in

Encyclopedia of Philosophy. Ed. Donald M. Borchert,

394-439, 2006.

Anderson, M., Anderson, S; Machine ethics: creating an

ethical intelligent agent. AI Mag. 28(4), 2007.

Bentham, J. Introduction to the principles of morals and

legislation. Blackwell’s political texts. Blackwell,

Oxford, 1789. Intr. de W. Harrison, 1967.

Bernard Williams. Negative responsibility: and two

examples. Utilitarianism: For and against, pages 97-

118, 1973.

Colin Allen, Gary Varner, and Jason Zinser. Prolegomena

to any future artificial moral agent. Journal of

Experimental and Theoretical Artificial Intelligence,

12(3):251-261, 2000.

Driver, J. The history of utilitarianism. In The Stanford

Encyclopedia of Philosophy, E. N. Zalta, Ed., summer

2009.

Lafollette, Hugh, ed. The Blackwell Guide to Ethical

Theory. Malden: Blackwell Publishers Inc., 2000.

Moore. J. H; The nature, importance, and difficulty of

machine ethics. IEEE Intell. Syst. 21(4). 18- 21 (2006).

S. Matthew Liao, A Short Introduction to the Ethics of

Artificial Intelligence In: Ethics of Artificial Intelligence.

Edited by: S. Matthew Liao, Oxford University Press,

2020.

Proposed Way to Inculcate Morality in Artiﬁcial Intelligence

321