An important first claim to make about my moral system is that it is objective. Its prescriptions are the same regardless of who is at the controls. Inputs to the system consist of empirical data regarding people’s degree of happiness or suffering. These data may consist of nothing more complex than the self-assessment of the people involved (how happy they think they might be if a certain action were performed). It might also include statistical data on the degree of happiness usually associated with a particular action. My purpose here is not to explore these sources of data in detail. I merely wish to emphasize that these data are, at least in principle, obtainable and objective.
Doctors often ask patients to rate their pain on a scale of 1 to 10. My suggestion is that we can extend this sort of scale to include the degree of happiness. I propose, then, that it makes sense to work with some sort of scale of happiness – call it the contentment index (CI) – that takes positive values for states of contentedness and negative values for states of suffering. Thus +10 on the scale would be great joy or pleasure while -10 would be extreme suffering. I assume that a person cannot simultaneously experience blissful happiness and excruciating pain, i.e. that he cannot have a positive and negative CI at the same time. In order for a suffering person to be content, her suffering must first be alleviated.
With the CI, we have a quantitative means for comparing different people’s levels of happiness. But CIs change over time, and people’s lives are not all the same length. How, then, do we compare CIs? To address this, let us consider some simplified examples of possible CIs (Figure 1).
In Figure 1 is plotted the contentment index of four hypothetical people, Allie, Bob, Connie, and Dmitri. CI is plotted along the vertical axis and time is plotted along the horizontal axis. Allie lives a relatively long and happy life, hence a graph that has positive values of CI. Bob is just as happy a person as Allie (the height of the grey box is the same for these two people), but Bob lives a much shorter life. Next is Connie, who also lives a short life but suffers constantly, hence a graph that has negative values of CI. Dmitri suffers with the same intensity as Connie, but lives for longer.
How should we rank these four lives? We have already decided that our goal is to increase happiness and reduce suffering. If this is the case, then a longer happy life is more valuable than a shorter happy life, because it contains a greater total amount of happiness. Therefore Allie’s life is more desirable than Bob’s. This is relatively uncontroversial (we seldom think twice before wishing people long and happy lives).
Conversely, if it is the case that suffering should be eschewed, then Dmitri’s life is less desirable than Connie’s because Dmitri suffers longer than Connie. Again, I think this is fairly uncontroversial. If most people had a choice between a long period of excruciating pain and a short period of excruciating pain, they would probably choose the latter. And people will often put a beloved pet down if it is experiencing great suffering, rather than allow it to continue in abject misery. (The reason we do not do the same with people has to do with consent, an argument I’ll develop in later chapters).
Figure 1. Four examples of contentment index over time. The scale on the right shows the utility of each example.
Finally, the lives of Allie and Bob are both preferable to the lives of Connie and Dmitri, because the former pair is happy while the latter is not. We therefore arrive at the ranking shown on the “utility” scale at the right hand side of Figure 1.
Although we reasoned the above rankings in a somewhat qualitative fashion, we were actually performing a very specific mathematical operation: We were calculating the grey areas in Figure 1. The scale to the right of Figure 1, then, simply shows the area carved out by each person’s CI curve (keeping in mind that areas below the horizontal axis are negative). I define the area under a person’s curve to be the “utility” of that curve. In summary, a large positive area has the largest utility (it represents a long period of great happiness), while a large negative area has the least utility (it represents a long period of great suffering).
Importantly, if the area defined by the graph is our metric for happiness, then a short but extremely happy life is equivalent in value to a long but only somewhat happy life.
CIs for Non-existent and Unconscious People
Some moral problems involve people who exist given one action, but do not exist given another. For example, deciding to have a baby will result in the presence of a new person, while deciding against having a baby will not. The existence of the baby is therefore contingent on the action chosen by the parents.
Comparing the utility of different actions must therefore be prepared to compare a non-existent person with an extant one. I take the approach of defining a value of CI = 0 for non-existent people. An analogy serves to demonstrate the thinking behind this approach. Consider a bowl containing water. Water, in this analogy, is happiness, while the bowl is the person who is happy. If we drain all the water from the bowl, the amount of water contained in the bowl is zero – a well-defined number. And if there is no bowl to begin with, then we can still say that the total amount of water present is zero. In the same way, I think it makes sense to talk about a CI of zero in the case of a person who does not exist. A CI of zero means that there is neither happiness nor suffering.
A similar approach can be taken toward people who are unconscious and therefore not capable of experiencing happiness or suffering. Using the water analogy such people can, like non-existent people, be said to represent the absence of a bowl with which to hold water. And since there is no bowl, the quantity of water must be zero.
My approach has precedent. In Practical Ethics, Peter Singer asks us to consider three possible universes:
1. The Peopled Universe, populated by 1 billion happy people.
2. The Nonsentient Universe, devoid of any life.
3. The Hellish Universe, populated by 1 billion people experiencing excruciating pain every minute of the day.
Singer argues that most people would prefer the Nonsentient Universe to the Hellish Universe, implying that it’s possible to compare a populated universe with an unpopulated one.
He makes a somewhat better argument by comparing the Peopled Universe to the Nonsentient Universe. To do so, he asks us to consider a series of intermediate universes between the Peopled and Nonsentient Universes, each with successively fewer life forms than the last. Just before we reach the Nonsentient universe, we see a universe containing, say, a single shrimp that experiences one passing moment of sentience before dying. Surely, Singer argues, it makes no sense to rank this single-shrimp universe on the same quantitative scale as the Peopled Universe, yet insist that the very next universe in the sequence, namely the Nonsentient Universe, be similarly ranked entirely differently.
A Nonlinear Contentment Index
Next, I suggest a refinement to the contentment index scale. I believe that suffering is a far more intense, overwhelming experience than happiness. Serious physical injury, such as torture, is far more intense and transformative an experience than, for example, the feeling of elation while listening to music. Put differently, our worst possible suffering is more extreme in its deviation from our baseline state than our highest achievable happiness. What this means is that the emotionally negative consequences of an action should be given more weight than the emotionally positive consequences of an action. Furthermore, a decrease in CI should be more heavily weighted if it involves negative values than if it involves only positive values (a decrease in happiness is not as “bad” as an increase in suffering).
This modification to the contentment index should be taken into account each time a CI value is assigned to an action.
Figure 1 showed simplified utilities over the full life spans of four hypothetical people. However, as stated previously, our aim is to consider which of several actions should be performed in order to achieve our moral goal. To answer this question, we do not generally need to consider people’s utilities very far into the future. We are only interested in the effects that specific actions are likely to have on utility, and these effects might be quite short-lived.
Let us consider a simple example of a moral problem with only two actions. The first action is eating an ice cream cone, while the second action is not eating an ice cream cone (doing nothing). Let us also assume that there is only one person in this problem, and that this person actually wishes to eat the ice cream. Her moral problem is this: Should she eat the ice cream or not?
Consider Figure 2. In the top panel we see what our friend’s CI is likely to be if she eats the ice cream. It is initially quite high thanks to the pleasure of eating the treat, but slowly decreases as these pleasant effects wear off. In the bottom panel, we see what her CI might be if she decides against eating the ice cream. It is initially quite low – she is disappointed at the prospect of passing up such a scrumptious dessert – but it eventually returns to a higher, more stable level as her disappointment dissipates.
The red vertical line in Figure 2 indicates the time at which the two CIs become indistinguishable. Beyond this point, the effects of the ice cream eating decision, whether for or against, have faded away. Consequently, there is no point in looking beyond the red line when making a decision about whether our person should eat the ice cream or not. Only the time interval preceding the red line is important, because this is where the differences between the two actions lie.
A comparison of the two CIs in Figure 2 indicates that eating the ice cream has greater utility (the area under the graph is greater), so our answer to the moral problem is that our friend should eat the ice cream. Of course, we could make the situation a little more realistic by including the long term suffering that consumption of the ice cream might cause in the form of tooth decay or weight gain. I’ll omit such complications for now, and simply note that they would not require a different methodology – they would only change the shape of the curves in Figure 2, possibly leading to a decision not to eat the ice cream.
Figure 2. Top: CI associated with the action of eating ice cream. Bottom: CI associated with the action of not eating ice cream. The two curves become the same at the time marked by the vertical dashed line.
Given the above demonstration, I now formally define action utility:
Action utility is the integral of the CI curve over time (it is the area under the CI curve). Importantly, action utility is specific to one action and one person.
Having defined the CI and its associated action utility, we are now one step closer to defining a moral calculus that will rank available actions on the basis of their utility, and therefore on their ability to fulfill our chosen moral goal. However, we have one more important preparation to make. We must determine which actions, beings, and time periods are relevant to a given moral problem. This is our goal in the next chapter.
Return to the previous chapter.
Return to the table of contents.