ALEX Institute / AI / AI Ethics · Module 2
Track C · AI Ethics, module 2. Turing Commons · about 13 min reading (2,554 words).
Chapter Summary
This chapter introduces the basic theoretical building blocks for learning AI ethics. We start with a brief metaethical motivation, looking at different positions in metaethics and reflecting on how these affect the way we do practical ethics. Additionally, we will delve into the some of the main normative theories in practical ethics: consequentialist theories, deontological theories, virtue ethics, and biocentric ethics. These concepts will be useful to paint the ethical landscape against which the sub-field of AI (practical) ethics has developed.
Learning Objectives
In this chapter, you will:
Practical ethics is a branch of philosophy, in particular a branch of ethics concerned with examining the aims and principles of moral behaviour and applying them to various problems of our everyday lives or real-world scenarios in general.
In this course, we will look at issues in practical ethics which arise from our use of AI systems and the prevalence of data in our lives. It is important to note that these harms are not exclusive to the use of AI systems, but can be characteristic of them. To do so, we will introduce some key concepts in metaethics and some of the most relevant ethical theories in use today.
Metaethics, as its name suggest, attempts to understand the various aspects (ontological, epistemological, semantic, and psychological) that make up the presuppositions and commitments of moral thought, talk and practice.
What does this mean? Metaethics deals with a variety of quite complex, yet relevant questions such as the following:
Clearly, none of these questions will have straightforward or easy answers, and we will not be dealing with them extensively in this course. However, in this brief introduction to practical ethics, we will look at different ways that some of these questions have been tried to answer.
Quote
What is morality? Are there moral facts about the world? If there are, can we know them? If there are not, how do we decide on right and wrong in our actions?
Moral realists, roughly speaking, accept the following two statements:
In practice, this means that moral realists believe that objective moral facts do exist, and there is only one "set" of these facts. Objective moral truths also imply that things like subjectivity, or culture are not relevant when evaluating whether an act is good or bad. Moral realism is, in and of itself, neutral on what it is exactly that makes something good or bad; it just is prepared to state that moral facts exist.
Of course, this does not mean that subjectivity or culture play no role in what people think is good or bad. However, the moral realist is prepared to say that those differences, and any human attitude over what makes something good or bad, is irrelevant when evaluating moral facts. That is, the reason why some things are good or bad do not hinge on whether people, culture, or society evaluate them as good or bad.
The existence of objective moral truths clearly begs the question: where do these facts come from? This is what is known as the 'grounding problem'.
One way of summing up moral realism is to say that epistemically they believe that "there are moral facts, in the same way that there are planets and spoons".
In procedural ethics, moral validity is bound by practices of giving and asking for reasons. In this way, a moral claim that is justified is one that can convince someone else by the “unforced force” of the better argument.
Legitimate moral decision-making thus requires persuasion through rational justification and compelling argumentation. This means that procedular ethics begins by trying to reconstruct the requirements for rationality. These requirements include assuming an impartial perspective, which considers the interests of all affected by an action equally, and making universalisable moral judgments that are applicable in all like situations.
Does this mean that procedural ethicists believe that moral facts exist? Yes and no. While processes of rational justification are supposed to yield a basis for valid moral claims, many procedural ethicists abstain from making substantive claims about moral reality. In their view, the proper role of moral theory is not to posit the fixed, and mind- or language-independent moral values and properties of the universe, but rather to facilitate and to normatively justify the continuation of ongoing rational dialogue.
Cultural Relativism says, in effect, that there is no such thing as universal truth in ethics; there are only the various cultural codes that derive from social approval, and nothing more. Cultural relativism challenges our belief in the objectivity and universality of moral truth.
Values and moral beliefs form an interconnected and coherent whole in a given cultural field or system of meanings, but this symbolic whole is relative to the historical background of the group it is derived from.
There are no objective moral standards that go beyond the locality of cultures, and hence, no universal standards that could help adjudicate between them.
Some examples of claims made by cultural relativists are:
Moral subjectivism takes relativism to a next level; instead of basing it on codes shared by whole cultures, they believe morality is an inherently individual matter. That is, what is morally right to do is up to the individual’s desires, wants, tastes, or preferences.
Therefore, in this view, moral language is not fact-stating language and thus does not contain mind-independent ethical truths. Instead, moral judgments are subjective expressions of feelings or attitudes, and so more like judgments of taste than statements of fact or rational justifications.
## Moral anti-realism The opposite of a moral realist is an anti-realist. As the name suggest, a moral anti-realist is prepared to declare either: 1. There are no objective moral facts. 2. There are moral facts, but they are not independent of human attitudes towards them.In contrast with metaethics, which is concerned with the nature of ethics, normative ethics is concerned with the content of ethical theories. Normative ethical theories attempt to give an answer to the question of what makes someone ethical, or what makes certain actions morally permissible. Ethical theories are varied and they come in different flavours. Here we will group them into four families of theories.
Consequence-based theories are normative theories which, as the name suggests, define whether an action is morally permissible in terms of the consequences it brings about. In simple terms, they define morality as a matter of maximising the best possible consequences in any given situation. This then begs the question, what are the best consequences to be maximised? Consequentialists have given different answers to this question.
One very influential flavour of consequentialism is utilitarianism, which specifies that the way we should rank consequences is by how much total (aggregate) utlity they bring about (where more utlity is better).1 This means that when confronted with a decision, utilitarianism will tell you to choose whichever action maximises overall utility. This theory's practical impact is not to be discounted. It has served as the basis for classical economics and is widely taught to and applied by economists today. Other examples of consequentialist theories are hedonism, act conseuquentialism, and rule consequentialism.2
In essence, what makes a normative ethical theory consequentialist is that, given a metric for what good consequences are (as well as a way of ranking better and worse consequences), it then states that in any situation, the morally approppriate way to act is to choose the action (among the existing possibilities) that brings about the best consequence.
Deontological theories in contrast, take a different approach. Morality is not to be measured by the consequences actions bring about, but instead morality dictates that one should follow a set of rules or principles regardless of what the consequences are. That is, the rightness of an action is determined by an agent’s application of some universal standard, rule, or maxim of rightness irrespective of its consequences and independent of the interests or ends of the agent whose conduct the standard, rule, or maxim is guiding.
The most famous proponent of deontological theories is Immanuel Kant, who was a firm believer in rule or duty-based morality, and who went as far as to define the maxim or principle to be followed in order to act morally which he called the Categorical Imperative. This rule roughly states that one must always act as if one’s actions could be willed into a universal rule that everyone in society follows.
In principle-based theories, morality is about following a set of rules or principles, and different theories will give different answers as to what principles one should act upon.
Virtue ethics is not concerned with how people should act in a particular situation, but instead on what kind of people they are. Its roots can be found in Confucius and Mencius in the East and in Plato and Aristotle in the West.
Instead of beginning with the question: “What should I do?”, it asks the question: “What sort of person should I strive to become in order to live an ethical life?”. Virtue ethicists emphasise that morality has to do with having a certain moral character which is achieved by cultivating virtues. A virtue is an excellent trait of character, and it comes in degrees.
These traits may derive from natural tendencies, but they must also be nurtured. Under these theories, a person who acts in a kind or benevolent way (these would be the virtues) does not do so because doing so would maximise the outcome of her actions (as a consequentialist would), or because it is her duty to do so (as deontology would advise), but instead because acting in that way cultivates said virtues.
The development of moral virtues through practice, discipline, and repetition is the purpose of the human form of life. For this ethical view to work, there must be some reserve of objective moral values available for agents to draw on to form virtuous habits and dispositions.
Some of the virtues stressed in antiquity but also justified in modern secular ethics: wisdom (ability to exercise practical reason to determine right action); temperance (ability to remain cool-headed and guided by reason instead of emotion); courage (ability to confront danger boldly and with self-assurance); justice (ability to act impartially and with fairness).
Biocentric ethics attempts to justify moral responsibility and the rightness of action not in a human-centered way but in terms of the intrinsic value of all life forms and ecologies. It treats “nature” (the living biosphere) not as an instrument or resource available to be used to achieve the purposes of human industry, but rather as an entity that makes a moral claim on us.
From a biocentric perspective, moral actions are those that preserve the flourishing and the diversity of all living beings and aim to secure the sustenance of the biosphere as a whole.
For the purposes of the rest of the course we will focus mostly on the first two families of theories.
Example
Let's take a classic example to help draw the distinction between consequentialism and deontology. Suppose an assassin comes to your door and asks if you have seen your friend Peter. And suppose you are indeed hiding Peter in your house as he is fleeing this very assassin. Should you tell the assassin the truth, or should you lie and tell him you don't know where Peter is?
From a consequentialist perspective, it is clear what one should do. Telling the truth will result in Peter dying and lying will result in him living. The first consequence is a lot worse than the second, so you should lie to the assassin.
However, from a deontological perspective it is not so obvious. If morality is about duty, and one of our duties is to be honest, then perhaps we also have a moral duty to tell the truth, even in this case.3
Most people would find the second conclusion morally abhorrent. This shows us that in some cases we seem to take a more consequentialist approach to ethics.
However, the converse can also be the case.
Example
Imagine you are a surgeon and Peter, your patient, is peacefully recuperating after an appendicitis operation. He should shortly be back home. And imagine that five people in need of different organ transplants come into the hospital. If these people do not get a transplant soon they will die. Suppose Peter is a match for all of these people. A strict consequentialist doctor might decide that we should kill Peter in order to save the other five via organ transplants. After all the consequences of one life lost and five gained are a lot better than one life gained and five lost.
Most people would find this conclusion morally abhorrent. A deontologist would reply that it is never permissible to use people as means to an end, since people are always ends in themselves. Therefore our duty to respect Peter's dignity and not use him as a means to saving the other five overrides any consequentialist considerations we might have.
The question as to what exactly utility is and how one goes about measuring it poses all sorts of complications which are beyond the scope of this introduction. ↩
For a list of different kinds of consequentialist theories see the Stanford Encyclopedia of Philosophy's entry on Consequentialism. ↩
There is a lot more nuance as to what a deontological theory would say in the example just described. However, this simplification helps illustrate a core difference between the two families of theories. ↩