II.1. Motivation

Here are some reasons to motivate a philosophical treatment of the idea of being good by doing good.

First, this idea is not uncommon outside philosophy. A number of sayings point in the direction that it is the goodness of a person's acts – rather than her intentions, asseverations, motivation, and so on – that should influence how we judge that person. For example, a common German saying is ‘Gut ist, wer Gutes tut’, which roughly translates as ‘He who is good does good’. Other proverbs pointing in that direction, though less clearly so, are ‘actions speak louder than words’ and ‘you are only as good as the last act you did’.

Second, the idea has some intuitive pull. Imagine, for example, a politician who leads his country into a war, because he believes it is the ‘right thing to do’. Not few would judge this person negatively, if he had good evidence that the war would have very bad consequences. So, despite good intentions and a conviction that he is doing the right thing, we might judge this politician negatively for doing something that has foreseeable negative consequences. Conversely, people whose charitable actions or scientific inventions have positive effects on the world are often judged more favourably as persons accordingly (even when we do not know their precise motivation or character).

Third, effective altruism has recently gained a lot of attention.Footnote ¹ Effective altruists are committed to using evidence and reason to determine the most effective ways to do good. For example, in the realm of philanthropy, they believe you should seek out a charity that is most effective at doing good rather than, say, one close to your heart. Similarly, effective altruists recommend choosing a career one has good reason to believe will do the most good, rather than, say, a career one is passionate about. In an op-ed piece in the New York Times, David Brooks takes issue with effective altruism. His charge: effective altruists confuse doing good with being good.

Effective altruists do not explicitly defend a theory of what makes a person good. But in much of their work, they do shift the focus away from motives and virtues towards how much good one does. One of the chapters of Will MacAskill's recent book on the topic is entitled ‘The Best Person Who Ever Lived Is an Unknown Ukrainian Man’.Footnote ³ We learn that said unknown Ukrainian man, a certain Viktor Zhdanov, effectively lobbied the WHO to eradicate smallpox, an intervention which did an incredible amount of good. Some effective altruists thus seem to think that being good and doing good are indeed much more closely linked than Brooks – and many others – believe. If independently plausible, such a view would help effective altruists fight off some criticism.

Finally, the approach of this article resonates with, but also differs from, an important tradition in moral philosophy. What we might call ‘virtue consequentialism’ holds that, first, being a good person is about being virtuous and, second, character traits are virtues, if they have good consequences. These broad commitments can be fleshed out in various ways allowing for different versions of virtue consequentialism.Footnote ⁴

David Hume's is probably the most famous version of virtue consequentialism. For Hume, moral evaluations are principally a matter of feeling. We judge a character trait to be a virtue, if it gives us the ‘pleasing sentiment of approbation’ upon surveying it.Footnote ⁵ A character trait typically gives us this sentiment – and this is the virtue consequentialist aspect – if it has good consequences in being either useful or agreeable.Footnote ⁶

More recently, Julia Driver has argued that ‘a virtue is a character trait that produces more good (in the actual world) than not systematically’.Footnote ⁷ Let us unpack this a little. First, virtues produce more good than not systematically, if they produce more good than not in normal circumstances. If a character trait does good by happenstance but not systematically, it is not a virtue. Second, to count as a virtue, a character trait need not maximize goodness. It is enough, if it produces more good than not.Footnote ⁸ Third, Driver remains neutral on what makes outcomes good or bad. Her theory of virtues allows for different views on axiology.Footnote ⁹

My approach takes a similar starting point as virtue consequentialism: fundamentally, good persons are such that they have a positive influence on the world. Virtue consequentialists fill in this idea by characterizing good persons with specific character traits that have good consequences (i.e. virtues). Unlike virtue consequentialists, I will discuss whether we can plausibly connect being good with good consequences directly, that is, without the need for virtues to act as ‘intermediaries’. If successful, we do not need to stipulate which character traits are relevant and in which sense they do good. Instead, we focus directly on how a person, taken as a whole, does good.

Let us consider another recent theory that bears similarities to my approach. In one of the chapters of his recent book, Philip Pettit argues that being good is about doing good; but not just any type of good, it is about providing goods that are modally robust.Footnote ¹⁰ For example, friendship is a modally robust good in the sense that being a good friend implies not only that one helps one's friend in the actual world but also that one would do so in relevant nearby possible worlds.

Pettit's theory of what makes character traits virtues differs from the aforementioned theories of virtue consequentialism. For Hume, a character trait is a virtue, because it is agreeable or useful. For Pettit, on the other hand, these thin benefits are important, but they do not exhaust the value had by virtuous dispositions. The underlying dispositions do not only causally produce other thin goods; they are themselves also constitutive of important rich goods. The core of Pettit's theory is thus a particular theory of the good according to which modally robust interpersonal values play an important part for overall goodness (and thus the associated notion of doing good). In this respect, it also differs from Driver whose theory of virtue is not tied to any particular view on axiology.Footnote ¹²

Again, my starting point is similar to Pettit's: being good is fundamentally about doing good. But there are differences.

First, my discussion of being good is meant to work without identifying specific individual character traits and dispositions – including virtues – as necessary for being good. The connection is directly between being good and doing good.

Second, my analysis does not make any assumptions about axiology. Of course, it is perfectly compatible with Pettit's theory of the good (and with many goods being modally robust). But my goal is to see whether we can find a plausible principle to link being good and doing good without making substantive assumptions about goodness or axiology. (I understand goodness, or the good, to be a property of outcomes or states of the world.) Accordingly, the goodness of outcomes could be welfarist, could be a function of different intrinsic goods, be preference-based or hedonist, and so on. My aim is to provide a structural discussion about how to derive judgements about how good a person is from judgements about goodness of outcomes. I will use examples in which, I hope, it is relatively uncontroversial which outcomes are better than others even in the absence of a theory of the good.

II.2. Desiderata

In this article, I pursue the following questions: first, what is the most promising rendering of the idea of being good by doing good? Answering the first question is necessary to answer the next. Does the most promising rendering of this idea provide us with a plausible principle of what makes persons good? And if so, does this principle tell us what makes a person good all things considered? Or is it only a plausible pro tanto principle, that is, a principle according to which a person is better the more good she does all other things beings equal?

How should we go about answering these questions?

Methodologically speaking, I will employ (a rough version of) John Rawls's reflective equilibrium methodology. Accordingly, a plausible principle should bring a set of relevant intuitions and relevant, more abstract moral beliefs into a coherent, reflective equilibrium.Footnote ¹³ Intuitions in this methodology are important, but they are not irrevocable. Contrary intuitions can be discredited in light of an otherwise very convincing theory.

More substantively, I suggest three desiderata that a plausible principle should fulfil.

First, we are here looking specifically for a principle that evaluates persons by how much good they do. The first desideratum is thus that a plausible principle should include all relevant goodness associated with the person being evaluated. Of course, the principle we aim to find is supposed to spell out precisely what that means. But, in the course of discussing different principles, I will repeatedly use examples in which we have clear enough intuitions about which good consequences matter for our evaluation. I will call the first desideratum the Relevant Goodness Desideratum.

Second, we should evaluate a person by only taking into account the goodness properly attributable to the very person under consideration. For want of a better name, let us call the second desideratum the Responsibility Desideratum. This is not meant to drive a clear dividing line between all possible sources for which a person is or is not fully responsible. This would be beyond the scope of this article and require discussions about moral luck and free will. Rather, this desideratum is meant to exclude those external sources most salient in our context. For example, an evaluation should not be based on a person's external economic circumstances and other people's decisions beyond her control. I will motivate such restrictions later by way of examples.

Lastly, our principle should account for, or cohere with, a number of more specific intuitions and beliefs. In particular, a plausible principle should have something to say in response to opposing intuitions and theories. For example, we might think that not only does it matter what a person does but also what she believes to be good or bad, right or wrong. A person's morally abhorrent beliefs (racist beliefs, for example) might make her a worse person over and above the beliefs’ bad consequences. Moreover, we might care about her motivation. Does she do good deeds only if they are in her self-interest or would she also do them if they were not? Does the person desire to do what is good? Does she do these deeds, as Kant prescribed, out of a sense of duty? A plausible principle will have something to say about these questions in response to other theories and cohere reasonably well with relevant intuitions. Let us call this the Coherence Desideratum.

I will discuss how far different principles (i) include all relevant goodness associated with a person (the Relevant Goodness Desideratum), (ii) fulfil the Responsibility Desideratum, and (iii) cohere well with intuitions regarding beliefs and motives (the Coherence Desideratum).

II.3. A few clarifications

I discuss principles that specify what makes a person good or bad. Some might object that engaging with the question of what makes a person good or bad is fruitless: for such judgements are either nonsensical or, less radically, reducible to judgements about actions, motives, moral worth or rules.

However, even without a detailed defence of the idea of ‘being a good/bad person’, we should not dismiss questions about it prematurely. First, such judgements are part of our everyday discourse. The idea of being a good person is very common in ordinary discourse, more common, indeed, than philosophical concepts such ‘moral worth’ or ‘virtue’. So, if nothing else, a philosophical inquiry will show us what our (imperfect) ordinary moral concepts are and how to render them more precise. Second, most people care about being a good person and thus have, at least in theory, an interest in finding out what it takes to be good. Third, while most (though not all) ethicists today focus on the rightness and wrongness of acts, this has not always been the case. For virtue ethicists, the moral quality of persons rather than actions takes centre stage.

Note that when I talk about ‘being a good person’ I mean for this to be a rather abstract higher-level concept. I do not take being a good person to be equivalent to having a good character or being a virtuous person. Whether being a good person implies having a virtuous character or something else (such as doing good) is an open question that needs to be determined through substantive moral theorizing.Footnote ¹⁴

Two further clarifications are in order.

First, the principles discussed here understand the goodness of persons in terms of betterness, that is, as a scalar notion. I will flesh this out by discussing rankings of persons in terms of the weak betterness relation ‘being at least as good a person as’.Footnote ¹⁵ The order is transitive, reflexive and antisymmetric. For simplicity, and to preserve a tighter connection with natural language, I will often talk about ‘good/bad persons’ and about ‘being good/bad’ as a shorthand for comparative judgements.

Second, I will mainly discuss how to rank persons in terms of betterness within a time span. The idea is that how good or bad a person is can change over time. Accordingly, we might say that a person strayed from the straight and narrow in her twenties but reformed herself into a comparatively good person later on. But we can also evaluate a person overall. For this, the time span under consideration covers her whole life. At some stages, I will omit the time span qualification to improve readability.Footnote ¹⁶

Let us now discuss substantive principles and see which precise version of the basic idea is most promising and, in a next step, see how plausible it is as a principle more generally.

IV.1. Goodness

A different approach judges a person by her actions instead of her existence. According to the first principle, a person is good to the extent that her acts bring about good.

I will here take ‘in total’ to mean the sum of goodness of a person's acts within T. Alternatively, we could use the average (or median) goodness of a person's acts within a time span. This would allow us to compare two persons with respect to two different time spans respectively. As both versions are open to the same objections, I will discuss the simpler sum version of this principle in which P and Q are compared with respect to the same time span T.

The Objective Act Principle fails to avoid the problems that beset the Existence Principle. First, a great many of Stalin's grandmother's acts were necessary elements in a chain of events that led up to her having a daughter. This daughter would later have a son that would grow up to be a tyrant. Many of the acts performed by Stalin's grandmother together brought about a lot of bad. Therefore, the Objective Act Principle would judge her to be a very bad person. Second, differences in abilities are still not factored into the evaluation of a person. Jane's acts bring about more good than John's acts, because Jane has many more abilities to do good. Therefore, she would be judged to be a better person than John on the Objective Act Principle. Third, an evil demon can turn the nicest of all people into very bad people by making his own actions contingent on their actions. Imagine, for example, that the evil demon decides to kill someone whenever Maria gets up before 10 a.m. Maria does not know about the evil demon. Still, her getting up before ten has terrible (unforeseen) consequences. Therefore, the Objective Act Principle will judge her to be a bad person. These judgements in the three cases above are implausible. The Objective Act Principle fails to meet the Responsibility Desideratum.

Maybe we should turn from objective to subjective forms of goodness. The notion of subjective goodness used here is evidence-relative. The evidence-relative goodness of an act is the goodness to expect relative to the evidence available to a person before or at the time of acting. This is different from objective goodness, because it is not the actual consequences of an act that matter. Evidence-relative goodness is also different from belief-relative goodness. Remember the above example of a politician starting a war, because he thinks it is the right thing to do. The criticism here is that this politician had sufficient evidence to believe that invading a foreign country would have bad consequences. Whether the politician believes invading would do a lot of good or bad is neither here nor there; what matters – in the principle discussed here – is what a person has evidence to believe. Henceforth, when I use ‘expected goodness’ of an act, I mean the evidence-relative goodness of an act.Footnote ¹⁸

The Subjective Act Principle gets round two out of the three problems. Because Maria did not have evidence that her acts – getting up before ten – would lead to so many bad things, we would not judge her to be a bad person for getting out of bed. The Grandmother Problem dissolves, because Stalin's grandmother did not have evidence that her actions would lead to an outcome in which her grandson would do so much bad.

However, the Subjective Act Principle does not get round the Abilities Problem. The expected goodness of Jane's action, although she runs a charity terribly, is still higher than that of John's actions. Thus on the Subjective Act Principle, Jane would be a much better person than John, because John has far fewer abilities to do good. So, while invoking expected goodness helps with some problems, it does not solve the Abilities Problem.

IV.2. Rightness

We have seen that using expected goodness helped with two of the three problems mentioned above, namely the Evil Demon and the Grandmother Problem. We will keep this element. But instead of the aggregate goodness of a person's acts, we will now focus on the expected goodness of individual acts. More specifically, the idea now is to see how often a person chose acts that did the most good. We can construct such a principle by invoking subjective consequentialist rightness: an act is right, if and only if there is no other act with higher expected goodness.

An immediate problem with this principle is, of course, that such a view assumes act-consequentialism about rightness. This section is thus not neutral about theories of rightness. But as the Rightness Principle is shown to fail, I will not have to defend this act-consequentialist assumption here.Footnote ¹⁹

What does ‘more often’ in the principle mean? We could focus on the cardinality of right acts:

As mentioned above, using expected instead of actual goodness avoids the Evil Demon and the Grandmother Problem. Focusing on consequentialist rightness seems to get round the Abilities Problem. The actions John is in a position to perform are lower in expected goodness than those that Jane is in a position to perform. Assume that John chooses acts with the highest expected goodness out of the set of available acts – those that are right according to subjective act-consequentialism – more often than Jane. According to the Rightness Principle, he is thus at least as good a person as Jane. This is the case even though John has fewer abilities and his acts are lower in expected goodness than Jane's acts.

However, while the Rightness Principle seems to meet the Responsibility Desideratum, it fails to fulfil the Relevant Goodness Desideratum. Compare two persons P and Q whose respective numbers of right actions are exactly equal. P’s wrong acts, however, are much worse than Q’s. Say, P gives the order to kill many people, whereas Q simply gives less money to charity than he could. On the Rightness Principle, P and Q will be judged equally good persons despite P’s actions having drastically worse expected consequences.Footnote ²⁰ The Rightness Principle thus fails to factor in how bad one's wrong actions are.

The act principles surveyed so far – both in their goodness and in their rightness versions – either judged persons by consequences not attributable to them (the Responsibility Desideratum) or failed to include all relevant goodness (the Relevant Goodness Desideratum).

V.1. Proportional expected goodness

We have seen that our principle should focus on expected instead of actual goodness to avoid the Grandmother and the Evil Demon Problem. But we have not found a principle that avoids both the Ability Problem and fulfils the Relevant Goodness Desideratum.

I suggest the following solution. We should ask: proportional to the expected goodness of a person's alternative courses of action, what is the expected goodness of a person's actual course of action? The idea behind this move is this. Jane, the rich heiress, could have foreseeably done a lot of good in her life. She has good evidence to believe that running her charity well would significantly increase goodness in the world. She also has evidence to believe that the life she actually chooses to lead will do much less good compared with possible alternative scenarios in which she runs the charity well. The suggestion is thus to compare the expected goodness of a person's chosen course of action with the expected goodness of alternative available courses of actions during a specific time span. A course of action is here understood to be a set of actions performed by an agent during a time span. Compare Jane with John. John decided to lead a life he had evidence to believe was high in expected goodness compared with alternative lives he could lead. His life was high in ‘proportional expected goodness’ yet relatively low in absolute expected goodness.

As seen earlier, by focusing on expected goodness we get round the Evil Demon and the Grandmother Problem. The proportionality element avoids the Abilities Problem. The principle thus includes all relevant information about goodness/badness whilst also fulfilling the Responsibility Desideratum.

Let me explain in more detail what I mean by ‘proportional expected goodness’. The main idea is that we first analyse the set of abilities available to a person over the relevant time span. (Again, ‘abilities’ are here understood widely such that a person's abilities determine which actions are available to a person.) Then we compare the expected goodness of the course of action actually taken with the expected goodness of the courses she foregoes. I will now explicate the different elements of this idea in three steps.

First, what is proportionality? One way to fill in expected proportional goodness would be to compare a person's actual course of action with the (expectably) best possible course of action available to her. We could, as a first approximation, use a ratio function, where ‘EG_A ’ stands for the expected goodness of the actual course of action and ‘EG_BP ’ stands for the expected goodness of the best possible course of action:

Say the expected goodness of the best possible course is 90 units of goodness and your actual, chosen course of action is 30. The proportional expected goodness according to M₁ would be one-third. The higher this number, the better one is as a person.

But there is a problem with M₁ . If the level of expected goodness is negative for the best possible course of action, then your level of personal goodness would increase the worse your actual course of action is (because both numbers would be negative). This is implausible. We can fix this problem by introducing ‘EG_WP ’, which stands for the expected goodness of the worst possible course of action. We first calculate the difference in expected goodness between the actual course of action and the expected goodness of the worst possible course of action. We then compare this amount to the difference in goodness between the best and the worst possible course of action:

One can think of the measure as follows. First, we want to know how much overall difference in expected goodness a person is able to make. We can think of the difference as the distance between the worst and the best possible courses of action (the denominator). Second, we then see how much difference is made in terms of expected goodness by the course of action actually chosen by the person. To do this, we ‘measure the distance’ between the worst possible and the actual course of action (the numerator). Then we compare these two quantities. As before, the higher one's number on M₂ , the better one is as a person. This measure has the advantage that it plausibly deals with cases in which EG_BP is negative.Footnote ²¹

Second, relative to what time should we fix the content of ‘expected goodness’? There are (at least) two ways to do so. On the first method, the Starting Point Account, the expected goodness of a chosen course of action over the whole time span is the goodness to be expected when choosing the first action of the whole sequence of actions. Say we want to calculate proportional expected goodness for person P during a time span T which spans time t to t + 7. We would then calculate proportional expected goodness according to the levels of goodness that P had evidence to foresee at t for the different combinations of actions between t and t + 7. The Starting Point Account, however, is very unforgiving. If we base our judgement only on the proportional expected goodness at the beginning, then making a mistake at the beginning of the relevant time span would make one a bad person throughout, even if one makes a strong effort to do good later on. Also, it seems a person should consider not only the evidence at the beginning of a course of action but also evidence that becomes available at later stages. The alternative is the Fresh Starts Account.Footnote ²² Here, we construct proportional expected goodness during a time span as a function of the expected goodness of the different combinations of actions at different decision-points. Calculating proportional expected goodness at various decision-points means that one can (sometimes) offset mistakes made at the beginning of a course of action. It also takes account of evidence about expected goodness that comes available at later times.

The Starting Point Account calculates proportional expected goodness for a time span T only once, namely at the beginning of T. The Fresh Starts Account, on the other hand, calculates expected proportional goodness at many different stages in T. I have argued for the latter. We thus need an account of cross-temporal aggregation for the various measurements of proportional expected goodness across time. I suggest we use the average proportional expected goodness across decision-points within a time span. This also allows us to compare persons with respect to different time spans respectively. (To keep things simple, though, I will continue using the above version which compares persons with respect to the same time span T.)

Third, when calculating the expected goodness of courses of action, the following question suggests itself: should such calculations include the probability with which a person is going to act certain ways in the future? Imagine, for example, that Francis wants to gauge the expected goodness of going to his friend Johannes's birthday party. Francis thinks a good party needs Reggaeton music. Johannes, however, only likes Electro. Francis cares deeply about Reggaeton and frequently gets into heated debates with Johannes. Johannes does not handle such debates well and his mood usually turns sour very quickly. I suggest that when we calculate the expected goodness of a chosen action, we need to include the probability with which a person will act certain ways in the future. So, when calculating the expected goodness of going to Johannes’s party, Francis needs to factor in that he is likely to get into an argument with Johannes.

However, when calculating the expected goodness of alternative courses of action, we do not include the probabilities with which a person is going to act certain ways. We compare a person's chosen action at a time and its likely future actions – so a probabilistically weighted course of action – with the best possible and the worst possible courses of action.Footnote ²³ When assessing Francis's case, for example, we calculate the expected goodness of the following courses of action: going to the party and not getting into a heated debate has the highest expected goodness; staying at home has the second highest expected goodness; and going to the party and getting into an argument has the lowest expected goodness. In calculating these levels, we do not include the probability with which a person will act certain ways. Why not? The reason is that if we did, a person's personality would be reflected in the expected goodness of the available courses of action. For example, a person who will probably act badly in the future would be judged much more mildly, because the range of expected goodness he is in a position to bring about would itself be restricted by his morally bad dispositions to act in certain ways.Footnote ²⁴

The measure presented here is relatively simple and more complex ones could be devised. I will leave some questions about measurement and aggregation unanswered, as the aim is not to solve all technical problems.Footnote ²⁵ Though this leaves us with some vagueness, our present measure is sufficiently precise to understand the implications had by the Proportionality Principle and assess its plausibility.

I suggest the Proportionality Principle plausibly meets the Relevant Goodness Desideratum – it includes all goodness relevant to personal evaluation – and the Responsibility Desideratum. But I will now argue that the Proportionality Principle does not by itself meet the Coherence Desideratum.

V.2. The Coherence Desideratum

Here are some intuitions and beliefs at odds with the Proportionality Principle.

First, some people might believe that a person with good (or correct) moral beliefs should be considered better than a person with abhorrent (or incorrect) moral beliefs – independent of the goodness caused by these beliefs.

By ‘all other things are equal’ I mean that all remaining criteria by which to evaluate persons together judge P and Q to be equally good persons. The Proportionality Principle conflicts with the Moral Beliefs Intuition. Imagine a person who has systematically wrong moral beliefs but who, for some funny reason, ends up choosing courses of action with high proportional expected good. We might think that such a person is less good a person than someone whose courses of action have the same proportional expected goodness, but who acts out of a strong moral conviction. The Proportionality Principle conflicts with such an intuition.

Second, another cluster of intuitions potentially problematic for the Proportionality Principle is about motivation. On the Proportionality Principle, being a good person is a function of the goodness of outcomes. Motives do not matter intrinsically. But many have the intuition that they should. There are many different intuitions about motivation. One such intuition is:

There is also the (broadly) Aristotelian idea that the virtuous person desires to do virtuous acts themselves (de re) rather than desiring to do whatever act is required for being virtuous (de dicto). Translated into terms of goodness, we might say:

A different intuition – a Kantian one – is that persons are better, if they act out of a sense of duty:

These intuitions are, of course, a caricature of Aristotelian and Kantian thought. Moreover, there is a potential tension between the Kantian and the Aristotelian Intuition. I shall not try to make these intuitions more precise nor try to resolve any potential conflict between them. The worry presented here is that the Proportionality Principle seems to ignore all such considerations, whatever their precise specification.Footnote ²⁸

I take it that the above intuitions are not prima facie unreasonable and cannot be shrugged off as irrelevant to the question under examination here. To meet what I earlier called the Coherence Desideratum in light of these opposing intuitions, we have (at least) two options: (i) we can grant the relevance of the above intuitions and thus either modify or renounce the Proportionality Principle; alternatively, (ii) we can try to explain away these opposing intuitions. Response (ii) would require showing that the Proportionality Principle can account for the spirit of the above intuitions without accepting that motives and/or beliefs matter conceptually for personal evaluation.

The ‘explain away reply’ runs like this. First, people with bad moral beliefs typically choose courses of actions that do little proportional expected good. Therefore, people with bad moral beliefs will typically – but not necessarily – be considered bad on the Proportionality Principle. If true, we can thus account for the spirit of the Moral Beliefs Intuition without assuming that a person's beliefs matter conceptually. Second, we can say similar things about people with good motives and/or altruistic attitudes. It is likely that a person who is willing to trade off her own well-being to do good will choose courses of action that are higher in proportional expected goodness than a very selfish person. The same might hold for the Aristotelian and/or Kantian Intuition.Footnote ²⁹

The ‘explain away strategy’ is made stronger by the Fresh Starts Account. Consider bad moral beliefs. For individual decisions, it is possible that a person who aims to do very little proportional expected goodness can fortuitously end up choosing courses with high proportional expected goodness. However, across many decision-points, it seems unlikely that people with abhorrent moral beliefs will happen to choose courses with high proportional expected goodness. The Fresh Starts Account will account for that, because it aggregates over proportional expected goodness across decision-points. Similarly, having morally bad (or neutral) motives might occasionally result in high proportional expected goodness. Across a very high number of decisions, however, systematic concurrence seems unlikely.

Does the ‘explain away strategy’ suffice?

One might first respond that the link between good beliefs/motivation and high proportional expected goodness is merely empirical and thus open to refutation. We can imagine circumstances in which not having an altruistic motivation will have high proportional expected goodness. For example, if the invisible hand of the market works well, people's egoistic motivation might often result in desirable outcomes – or so the idea. At the very least, one might argue that the ‘explain away strategy’ relies on empirical assumptions the defence of which cannot be satisfactorily done from the philosophical armchair.

But there is a different, more systematic problem for the ‘explain away strategy’. Consider:

If the prisoner does A, he will get a ‘perfect score’ on the Proportionality Principle during that time span. But it seems counterintuitive to say that the prisoner is a good person overall. I think Racist Prisoner shows that the Proportionality Principle cannot be an all-things-considered principle. Instead, the Proportionality Principle is just one principle amongst others. The prisoner's contribution to overall good is too insignificant to justify any strong judgement about him. Moreover, we still intuitively judge the prisoner to be a (comparatively) bad person in this situation for reasons other than his proportional goodness/badness. To account for this thought, I suggest, we need to add further principles.Footnote ³⁰ The Proportionality Principle is, I submit, a pro tanto principle. The Proportionality Principle does not give us necessary and sufficient conditions for betterness rankings. We need further principles that capture the idea that motivation or beliefs matter beyond their consequences (these principles spell out the ‘other things being equal condition’).

This version of the Proportionality Principle is only a pro tanto principle, which alleviates some of the counterintuitiveness of Racist Prisoner: the prisoner is ‘slightly better’ in virtue of doing some proportional good during T, but he is still not a good person overall.

But now one might object that making the principle pro tanto is not enough. The Proportionality Principle still makes judgements on persons who have very little opportunity to do any good/bad. Is this not a problem in itself? For three reasons, I do not think it is.

First of all, note that the Proportionality Principle will not be defined in extreme cases. If the best and the worst possible courses of action open to a person have the same expected goodness, our measure M₂ has zero in its denominator and is thus undefined. That our measure is not defined in such situations is plausible. For when a person's behaviour makes no axiological difference whatsoever, it does not make sense to say that a person ‘does good/bad’. That means the Proportionality Principle only gives us a partial and not a total order.

Second, in situations like Racist Prisoner, a person can make a tiny difference in terms of expected goodness. On the pro tanto Proportionality Principle, this tiny difference still influences our judgement of how good a person is overall. One plausible response here is to say that the strength with which the Proportionality Principle influences judgements of how good a person is overall can vary across situations. Imagine we are comparing persons P and Q in terms of how good they are as persons overall. P performs higher on the Proportionality Principle than Q. But Q scores higher on the other principles (those to do with motives and so on). Who is better overall? This depends on two types of ‘quantities’. The first one is the difference between P’s and Q’s score on the Proportionality Principle. The second is what we might call ‘inter-principle strength’: in determining how good a person is overall, we need to balance different principles with each other. When determining how good a person is overall, the weight we give to one principle in comparison to another is its inter-principle strength. One option here would be to say that the Proportionality Principle's inter-principle strength is constant across all situations and persons. Instead, however, I think we should say that inter-principle strength can vary across situations and persons. For example, for a person with next to no opportunities to do good or bad, the Proportionality Principle would weigh in less heavily than for someone who has ample opportunity to do good or bad. For example, in Racist Prisoner, the Proportionality Principle will simply be a less important consideration than it is for people with ample opportunity to do good or bad.

Finally, Racist Prisoner is a fanciful example. As long as we take sufficiently long time spans, nearly all real-life examples will be such that a person can make significant axiological difference. For example, most people are in a position to benefit or harm another sentient being in some significant way. So, for most real-life cases, a drastically varying inter-principle strength might not be necessary.Footnote ³¹

The idea that ‘good persons do good’ is powerful and intuitive. Moreover, in many cases the Proportionality Principle gives us the intuitively right ranking. The Proportionality Principle spells out one important aspect of being a good person. However, as was seen in Racist Prisoner, it also has limitations. The Proportionality Principle is only one principle amongst others of what makes a person good.