Part 5 – Moral Decision Theory

Part 5 Prelude

Moral Decision Theory

Intuiton

To validate a position of moral philosophy, people will often refer to their own intuition. If the moral code requires that we perform Action X but our intuition suggests that we should perform Action Y, this is seen as evidence that the moral code is incorrect. The proponents of intuitionism may therefore use their position to argue against MDT.

https://plato.stanford.edu/entries/intuitionism-ethics/

The use of intuition can be useful in probing and questioning a position, in searching for inconsistencies, and in revealing potential gaps or shortcomings. This is the approach taken in Part 5 below. However, it doesn’t stand to reason that where a moral philosophy is unintuitive that we should trust our intuitions. If this were the case, we would simply agree that “We ought to act in accordance with our intuition” and cease any further thought or reflection.

In his book “Thinking, Fast and Slow”, Daniel Kahneman described two modes of thought, instinctive and deliberative. Both can be used to identify which action to take. The first, known either as thinking fast or as intuition will be used in the great majority of decisions. We simply don’t have the time, or the patience, to deliberate upon every choice we make.

There may be a conflict between our intuitions and rational deliberation. Either approach has its benefits, but importantly, either system is prone to error. Whichever system of thinking an agent actually uses to make a decision doesn’t have any bearing on what decision the agent should make. Only logic is objective, and only logic can be used to identify the epistemic modal facts. To put it simply, we can use logic to refute our intuition, but we can’t use our intuition to refute logic.

Part 5 Moral Decision Theory

In the Extreme

In this part, I will stress test Moral Decision Theory. We will consider certain edge cases or extreme situations which may highlight the challenges of the approach. If there are any gaps or inconsistencies with Moral Decision Theory, we may find them by looking at situations which are rare or unlikely.

We can check whether MDT is unable to provide an answer to the question of what decision an agent should take. We may also identify those situations where MDT prescribes an action which is inconsistent with our intuitions. At this point, one will either lose faith in MDT, or will “bite the bullet”.

It’s also worth noting that for some of the situations below, I have provided the response which I understand to be the most logical. The application of the theory in extreme cases may be very complicated, and I do not claim to be a perfectly rational being. Different approaches to these situations may be considered which are consistent with Moral Decision Theory.

If the reader believes that an alternate rational response can be provided to an extreme situation, there is simply a disagreement about the application of MDT. However, if the reader finds that no rational decision can be made in certain extreme cases, they may find this to be an objection to the theory itself.

5.1 The Mixed Perspective

In section 3.2.1 we discussed the situation of Tim and his glass of water. Tim is thirsty, and on the table in front of him, he perceives a glass of water. That “Tim should drink the water” is an unconditional prescriptive epistemic modal fact.

Now, suppose that while Tim was momentarily out of the room, a psychopathic ninja sneaked into his house and contaminated the water with an odourless and tasteless deadly poison. The ninja then disappears without a trace. On his return, there is no way for Tim to know, or even suspect, that this event has occurred.

Despite the actual fact that the glass now contains poison, it is still true that “Tim should drink the water”. This obviously seems counter-intuitive. Should Tim really drink the poisonous water?

The reason that this statement seems illogical is because it involves a mixed perspective. It expresses an epistemic fact, but because we have access to additional information, we assume that the sentence refers to the actual facts of the situation. We find it difficult to separate these two perspectives – epistemic and actual. It might be more intuitive if we rephrase the sentence as “Given the information available to him, it is reasonable for Tim to decide to drink the water.”

That this sentence seems unintuitive is not an argument against epistemic modal facts. It simply points out that expressing them can sometimes lead to misunderstanding. This occurs when different people have access to significantly different information. We need to be clear with our language in such cases.

5.2 Relative Objectivity

The next set of situations examine the cases where our use of language expresses both our emotional attitude as well as a statement which aims at objective truth. It is important to disentangle these two.

5.2.1 The Second Best Option

According to Moral Decision Theory, to say that “The agent should perform Action X” is equivalent to saying “Action X is expected to result in the best range of subsequent events”. But this does raise the question of the second best option.

Suppose that Larry has $10 in his pocket, when he passes a charity volunteer on the street. Larry already knows about this charity, and understands that the cause is important and that the organisation is effective. According to Moral Decision Theory, he should donate his $10.

But does this mean that he should not donate $9? If he were to donate only $9, would this be an immoral act?

Ultimately, the use of the word “should” is a matter of convention. We simply use language to communicate certain ideas, and as long as we are clear, then it doesn’t really matter how the word is used. Remember, the term “should”, when used to express Unconditional Prescriptive Epistemic Modal Facts does not imply any normative force or obligation.

The facts relevant to Larry’s situation is that it is best to donate $10. It is better to donate $9 than $8, but is not as good as if he were to donate $10. The use of the word “should” will depend on what we are trying to express, in the specific context.

5.2.2 The Status Quo

Something unique to Moral Decision Theory in relation to many other approaches to moral philosophy is that there is no default situation. We consider the subsequent events of our decisions, rather than the consequent events of our actions. Inaction is treated in exactly the same way as action.

Suppose that Laura regularly visits her local bar. The two bartenders, Bob and Mike both know that Laura has a severe allergy to pineapple. On one occasion, Bob notices that Mike has knowingly poured pineapple juice into Laura’s drink. She doesn’t notice and is about to take a first sip.

Bob has a decision to make, he can either do nothing, or stop Laura. A very likely potential subsequent event, if he decides to do nothing, is that Laura will have a life-threatening reaction. From this we can conclude that Bob should interfere with the situation. Objectively, he would be wrong to do nothing.

It is also objectively wrong for Mike to have put pineapple juice in the drink. Our intuition is that somehow Mike’s action is morally worse, than would be the case if Bob didn’t intervene. However, Moral Decision Theory offers no such distinction. For both agents, there is a very negative subsequent event which is likely to occur following the choice they have made. The moral facts of each situation are similar, in that both have made the wrong choice. Directly causing harm, and not intervening to avoid harm, are not inherently different.

Mike may have had malicious intentions, while perhaps Bob was simply afraid to intervene. Because of this, our negative feelings towards Mike may be more intense than towards Bob. One may be described as evil, while the other is only thought of as cowardly. But these statements reflect our subjective attitude, not objective fact.

This has implications for how MDT will be applied in a range of situations, both real (cases of abortion or euthanasia) and theoretical (the classic trolley problem, or Peter Singer’s child in the shallow pond). These will be investigated further in the future.

5.2.3 Unforeseen Consequences

Another situation often discussed when considering morality relates to unforeseen consequences. Suppose Adam is driving down a road at a safe speed, is concentrating on the road, and has taken every reasonable step to ensure the safety of his vehicle. Unfortunately, a child runs onto the road from behind a parked car. Despite his quick reaction, Adam can’t break in time, and hits the child.

This might be considered to be an unforeseen consequence of Adam’s decision to drive. But of course, whenever someone drives there is a small chance of having an unavoidable accident. This event may be unforeseen but it is not unforeseeable. According to Moral Decision Theory, all potential events are relevant for decision making. Whether they affect the correct choice will depend on the likelihood and severity.

A yet to be written section will deal in more detail with the concept of Moral Luck. In the meantime, it is enough to reiterate that extremely unlikely but extremely negative consequences are not to be judged in hindsight, but in relation to the information available to the agent at the decision point.

5.2.4 Forgetfulness

The foundation for epistemic modal facts is information, most of which is available in the form of memory. This raises a question about forgetfulness.

Suppose you notice that you’ve run out of milk. Later in the day, you walk past a grocery shop, but you don’t think to go in and buy some milk. You had forgotten that you had run out. Is it an epistemic modal fact that “You should have bought the milk”.

We can imagine a more serious situation. You visit a foreign country for the first time, and read in your guide book that the phone number for emergency services is 789. During your trip, a passerby has a heart attack while no one else is around. Unfortunately, you are struggling to remember the correct phone number for the emergency services. Is it true that “You should call 789”?

In these cases, it clearly wouldn’t true if you didn’t have the information at all – if you weren’t aware that the milk had run out, or if you had never learned of the emergency number. On the other hand, if these pieces of information were at the front of your mind, then it clearly is true.

In the first instance, if one was to ask, as you passed the shop “Do you need any milk?” you would have answered “Yes”. In this case, the information is available to you, so it is reasonable to claim that you’ve done the wrong thing by not entering the shop.

In the second case, the information cannot be recalled immediately, so it is not available to the agent. In response, you would have a decision about whether to spend time trying to remember, or to accept that you don’t know the number and assist in another way. This correct decision here would depend on the information you have relating to your own ability to recall forgotten information.

5.3 Dealing with Infinity

5.3.1 The Best of an Infinite Bunch

A challenge for any imperfectly rational agent hoping to implement MDT is that in the majority of situations they will have an effectively infinite number of options. Suppose Zainab is deciding how to donate her surplus cash. She could essentially give it to almost any mix of causes or people. According to MDT, there is a best option, but does this even make sense among an infinite number of options?

Just as when we count, there is always a larger number, perhaps for certain decisions, there is always a better option to consider. The response to this problem is to realise that there is a limit to how good our options can be.

Suppose we wanted to know the maximum life span of a fruit bat. The oldest ever observed is 33. We could perhaps imagine that one could live to 34. Perhaps 34.5 or even 35. But that doesn’t mean we can continue to increase the age forever. It’s clear that a bat cannot live to be 1,000. Instead of considering the maximum life span as an actual number, it can be considered to be a limit against which the age of lucky old bats approaches.

So it is with our best option. With each conceivably better option available to an agent, the marginal difference will diminish. As this limit is approached there comes a point where there is no discernible difference. Remember that the best range of expected consequences is determined by the information available to the agent.

Our information regarding experience is not quantifiable in terms of numbers or statistics. The comparison and ranking of ranges of potential subsequent experiences is limited by the agent’s ability to perceive the minor differences between them. In reality, one experience might be marginally better or worse than another. For example, a pleasurable experience that lasts a fraction of a second longer, or a burning pain which fires one additional neuron. But since the agent can’t conceive of such differences, the potential experiences according to the agent’s information renders them as equal.

Suppose you were buying a ticket to watch a football match. You have a choice between seats F17 and F18 in Block L and seat D4 in Block C. When comparing Block L and Block C, you can reason about how the view might affect your experience. You can weigh up the potential benefits against the potential costs. However, you won’t be able to make a similar analysis between the two seats within Block L. As they are next to each other, the information you have which determines the potential experience is identical for each. Neither option is better than the other.

So it will be in every situation, that as options approach the hypothetical best option, the agent becomes indifferent about which selection should be made. We may not be able to use logic to identify the single best option, but instead can identify an option which is indistinguishable from the logical limit.

5.3.2 Predicting an Infinite Chain of Events

On a similar theme to the point above is the consideration that our actions can result in a chain of events which go far beyond our ability to predict or interpret the outcomes.

Suppose that a parent is wondering whether to work late one evening, or to go home and help their child with homework. Perhaps working late will increase their chances of getting put onto a new project team. This could lead to career advancement. This could propel them to new opportunities and a higher salary which could mean a larger transfer of wealth to their child in the future. Alternatively, helping their child with homework could encourage a positive attitude towards learning. This may give the child a desire to do well at school, leading them to find more fulfilment and success in their own career. These two options will have only a marginal effect on the distant future, but the chain of potential events is slightly different depending on which option is chosen.

It is important to remember that the further from the decision point a subsequent event is, the less likely it is to occur, and the less our actions will influence the event. In the example above, staying late one night may increase the agent’s career prospects by a very minor percentage, relative to the alternative. The same is true for the immaterial additional impact that one night will have on their child’s behaviour. Instead, the deciding factor in a decision will be those subsequent events which are more immediate. The far-off potential consequences are irrelevant in the grand scheme of things.

We might also wonder whether two infinite sets of potential experiences are comparable.

Just because some range of possibilities is conceivably infinite, doesn’t mean that it cannot be compared to another. Suppose I were to plant an apple tree and an oak tree. There are an infinite number of shapes that sizes that each tree may potentially grow into. Regardless, it is a fact that the oak tree is likely to be larger than the apple tree. It is the same with a series of subsequent events which may follow from each course of action, one range will be better than another.

5.4 Extreme Scepticism

When agents make decisions, they must form beliefs about the world around them, and about how their actions will influence potential events within that world. But what if their perceptions of that world are illusory? What if the agent is simply a brain in a vat? What if the laws of nature upon which previous events occurred suddenly changed such that our predictions are rendered irrelevant?

These situations can be grouped together, along with many others which could be imagined, under the heading of scepticism. How are we to make logical decisions under such circumstances?

Before considering how these extreme possibilities affect our decision making, we can consider a more practical example. Suppose you and a friend are sitting on the sofa. On the coffee table in front of you is a pile of playing cards. Your friend offers you a small bet. You will guess whether the top card is red or black. If you are correct, you win the bet.

If this were a normal deck of cards, your odds of winning the bet would be 50%. However, you know that your children often play with multiple decks of cards, and that this pile of cards is not a standard deck. It could have any number of red or black cards. For all you know, there could be 5 red cards and 35 black, such that betting on red would seem foolish. However, based on your information, the pile of cards could equally favour red.

Although there is a good chance that the deck is not standard, we can essentially ignore that possibility. Based on the information you have, the probability of each colour is still 50%. The decision is unaffected.

Now suppose that you have a decision to make about whether to spend your Friday night volunteering to help vulnerable people or watching TV. You recognise that there is a chance that you are actually living in a simulation. Perhaps, when Friday night comes around, you will actually find yourself unplugged from the simulation. Or perhaps everyone in the simulation will suddenly be given superpowers. Perhaps your legs will turn into wheels, or your hair will become spaghetti.

While all of these occurrences are theoretically possible, you have no information to suggest that any is more likely than any other. Because of this, such potential outcomes have no bearing on the decision at hand.

Looking at this another way, we can consider that there are two general possibilities. Either the information available to us is reliable in making predictions, or it is not. We can define the probability of having reliable information as X. The probability that our information is unreliable, for whatever reason, will be (1-X).

The value of X is impossible to determine, because we have no information upon which to make such a judgement. In this sense, there is no actual probability, since we have defined probability as the reliant upon information. For argument’s sake, we can label it as “uncertain probability”, essentially representing any value greater than 0 and less than 1.

When making a logical decision about how to act, we need to compare the potential subsequent events of each available course of action. Suppose that according to the available information, taking Action A has a 50% chance of leading to Event B. To take account of our scepticism, we would then multiply the probability of occurrence by X. So Event B has a likelihood of 50X%. Of course, we are multiplying every single event by X, for all possible events and all possible courses of action. If a particular option is best without regard to X, then it will still be the best option after we include X.

Our scepticism has no bearing on our decision, because it cannot inform our choices one way or the other.

5.5 Are You Experienced

We can probe the prescriptions on Moral Decision Theory by considering some unusual examples of experience. For prescriptive facts to be unconditional they should apply to all sentient beings, in all possible worlds.

5.5.1 Pain Asymbolia

There is a condition known as “Pain Asymbolia”, in which someone can feel pain but they do not find it to be unpleasant. This may be because of some injury or drug which is interfering with the usual working of their brain. If an action was likely to cause someone pain, which they didn’t care about, should the agent avoid this potential experience?

It would be hard to empathise with such a person, having never experienced this situation before. Our information is limited, but that doesn’t mean a logical decision cannot be made. The agent would need to consider both the pain and the response to that pain as part of the potential experience which may occur following their decision. Clearly the pain itself should still be avoided if possible, but is preferable if the alternative was pain being experienced by someone without this dissociation.

5.5.2 The Experiences of Others

Suppose that an alien from the planet Lazhak really enjoys the sensation which a human would experience when stubbing their toe. For whatever reason, the stubbing of a toe is beneficial for Lazhakians. As a result, their evolution has resulted in them favouring the associated sensation.

According to Moral Decision Theory, an agent should seek to avoid or reduce those potential experiences which according to their information are negative. Suppose that a human had a decision to make such that Action A would result in a Lazhakian stubbing their toe, and Action B would not. Based on the human’s memories of such experiences, they would be correct to choose Action B.

However, if they had the information that Lazhakians actually enjoyed that feeling, this would be sufficient to change their decision. With that new piece of information, they have learned something about the disposition of the affected individual. This changes the potential experience from a negative to a positive one.

5.5.3 The Remembered Self

Experiments have shown that people have different perspectives on experience whether evaluated in the moment or in retrospect. This has implications for decision making given that the information provided by our memories is not an accurate representation of actual experiences.

It also should affect how we make decisions which may affect future experiences. For example, a negative experience may have echoes of traumatic memories. The experience of remembering something negative, is itself a negative experience. This means that all of those potential experiences, including reminiscing, should be considered when making a decision.

https://www.youtube.com/watch?v=4piJGDft3zs&ab_channel=LexClips

While the separation of the self into two parts may cause challenges, it does not invalidate Moral Decision Theory. We should simply aim to take into account all forms of experience.

5.6 Aesthetic Choices

According to Moral Decision Theory, for all decisions, the choices available can be ordered from best to worst using logic. This raises questions about decisions which are generally considered to exclude rational thinking, perhaps almost entirely.

Suppose that an artist is considering what shade of red they should use in a particular painting. According to MDT, there is a correct answer to this question. This seems unintuitive. We generally feel that thinking inhibits the free expression which is required for art and creativity.

The simple response to this question is to say that the artists should make the decision to act without thought. In this case, the expression of emotion or mood is itself an action. The correct decision, the one which will bring about the best set of mental events, is to pursue such an action.

5.7 Taking Risks

Probability and expectation are key to rational decision making, and therefore are key to Moral Decision Theory. This raises questions about risk and risk tolerance.

Suppose that Li Wei has a choice between Action A and Action B. According to the information available, Action A is 100% likely to result in a positive experience, while Action B is 50% likely to result in two similar positive experiences. According to Moral Decision Theory, neither option is better than the other. Li Wie should be indifferent.

We can manipulate these percentages further and at a certain point will find a decision which readers will feel differently about. Suppose you had a choice between a 100% chance of a negative experience, or a 1% chance of 100 such negative experiences. Perhaps some readers will intuitively prefer the first option, while others will prefer the second. This will depend on your risk appetite, which is subjective.

It may seem that Moral Decision Theory is missing something if it cannot account for such an imbalance. The response is to consider that risk or uncertainty is not just an element of decision making, but that it also leads to feelings in individuals. Some people feel terrible anxiety as a result of uncertainty, while others get an adrenaline rush from taking risks. These potential experiences need to be taken into account just as any other. Information about the risk tolerance or appetite of the affected individual will inform how they may respond in a particular situation.