The Cause and The Omission
ìåøéàï ÷øèñ úåàø ùðé ú.æ. 322029595
2 feb: am adaugat o corectura: Cazurile de early preemption POT fi cazuri in care un delayer e efectiv. Nu neaparat ca sunt.
Dar late preemption sunt neaparat cazuri in care un delayer e preemptat.
am mai adaugat un paragraf esential: asta
This paper has two purposes. The first is to elucidate the concept of The Cause of an event. The second is to provide a place for omissions as causes. It is a truism to say that no event happens as a result of a single cause. To my knowledge, analyses of causation are concentrated exclusively on analyzing the concept of a cause. On the other hand, common-sense intuitions often distinguish between The Cause of an event and a cause of it. Common-sense also provides for a metric of causation: an event A may be a cause of another event E, but not so much as event D – which is more of a cause of E. This metric of causation is important in moral contexts, where responsibility for an event is shared by several human agents. In most such contexts, it is true that several agents caused an event; but it is not true that all of them are equally responsible for it. The thesis which I will develop is that the metric of causation present in such judgments can be analyzed (in a probabilistic way) in terms of the closeness of a cause to The Cause. The Cause of an effect is, intuitively, that cause which causes it the most; which has the most significant contribution to the occurrence of the effect. I will use the majuscules (The Cause) in this paper in order to avoid confusion with the anaphoric the cause understood as that among the causes which was mentioned above; but I will try to avoid using the words the cause in this sense.
The second part of my essay is independent from the first one. In it I will defend the admission of omissions as causes of an event. The admission of omissions as causes is an ugly consequence for those who tacitly think that we are responsible for all the events that we cause: they think that accepting omissions as causes leads to the unwelcome conclusion that we are responsible for every act we cause by our omissions. (This is not the only reason why omissions are not accepted as causes: another, more ontological reason is that omissions are disjunctive events). We cause by omissions a lot of events, specifically all the events that we would have been able to prevent in a way or another, or perhaps even those events that we had a slight chance of preventing if we really wanted to (including the death by hunger of those people to whom we did not donate food, or the wars that were started by mad dictators whom we did not try to kill): since we do not want to feel responsible for all these events, an easy and seemingly convenient solution is to deny that we cause anything by our omission.
I want to reject this solution. The reason is that we cause by our acts a lot of bad things, perhaps as many and as bad as those whom we cause by our omissions. Lewis gave the grotesque example of a young man who works in a town; the director fires him, that`s why the young man moves to another town, where he meets a rich woman whom he marries. As a result of their marriage, they have a child; as a result of that child`s birth, the child will die some tens of years later. In this example, the events that were mentioned form, in their order, a causal chain; and even leaving aside transitivity of causation, the last event would not have happened in the absence of the director`s hiring the young man. It would be, however, absurd to consider the director responsible for the death of the child of the man whom he fired.
The actual incentive for writing this essay was to defend the thesis that the distinction between killing and letting die is of no intrinsic moral significance. A popular argument against this view is the following: "If letting die were ceteris paribus equally moral (or immoral) as killing, then we would be guilty for murdering all those whom we let die – for example people who die of hunger in Africa". More generally, the abolition of the moral significance of the distinction between acts and omissions leads to unbearably high moral standards: we would be guilty for all the bad things we cause by omission. (since presumably we cause more things by omissions than by actions).
My thesis is that the above argument rests on a mistake; the mistake is to think that we are responsible for all the events that we (actively) cause. In the light of Lewis` example, it is obvious that the objection too many moral obligations that is formulated against the thesis that omissions are not morally different from acts, can work equally well against the thesis that we are responsible for everything we cause by our acts. The truth is, in my opinion, that we are sometimes responsible for the consequences of our acts and sometimes responsible for the consequences of our omissions; we are sometimes more responsible and sometimes less responsible for these consequences, but we are not more responsible for the consequences of our acts than for the consequences of our omissions. We are more or less responsible because causation comes in degrees. We are the more responsible for something the more of a cause our act, or our omission, is. I will argue that, if a distinction is to be useful for ascription of moral responsibility, this distinction is not that between acts and omissions, but that between causes closer to being The Cause and causes farther from being The Cause.
I. FOUR CRITERIA FOR BEING THE CAUSE
If determinism, at least weak determinism, is true, then there are sufficient conditions of causes, and the conjunction of the causes of an effect gives its effect probability 1. If indeterminism is true, then the conjunction of causes we can specify raises (ex post facto in an unreversed way) the probability of the effect. I am not concerned with the question whether determinism is true, i.e. whether there are sufficient causes of effects. I will treat determinism as an ideal case of a cause raising the probability of the effect.
My purpose is to clarify the concept of The Cause as different from that of a cause. I will offer a few hints towards what I think is a fulfillment of such a task.
1. Principle of Sufficiency: The Cause is that among the several causes of an event which is the closest to being a sufficient cause. This means that The Cause is that cause whose occurrence, outside the sufficient conjunction of all the causes, is the most often followed by the occurrence of the effect (and is a cause of it). It may be tempting to think that the formula which captures this intuition is:
(For any a) C is The Cause of E if p(C&E)- p(C&~E) > p(a&C)-p(a&E)
where a ranges over the causes of E.
This formula has a very simple meaning, namely that The Cause is that among the causes which appears the least often without the effect. If C is a sufficient cause of E, then p(C&~E) =0, since p(C&E) = p(C). However, this formula does not capture the meaning of being the closest to a sufficient condition of the effect, since it may be possible that C (the sufficient condition) be not necessary and appear pretty seldom at all, albeit whenever it appears it is followed by the effect, while another, non-sufficient cause D, appears very often and some of the times when it appears it is followed by the effect. In such a situation,
P(D&E)-p(D&~E)
may be very high, not because p(D&~E) is low, but because p(D&E) is high. We do not want, however, to exclude rare events from the candidates for The Cause.
That’s why the formula which I propose for capturing the intuition of being the closest to a sufficient condition of the effect is
C is The Cause of E in the sense of being a sufficient condition of it if
For any cause a of E, p(E/C)>p(E/a)
Ideally (in a determinist world), if C were a sufficient condition of E, then p(E/C)=1
Davidson (Causal Relations) made the remark that the more detailed our description of a cause is, the higher our chances that it be a sufficient cause. (Another way to formulate this remark is to say that the more fragile a cause is, the closest it is to being a sufficient cause). This is trivially true, since a cause that is very fragile is an event that can occur very rarely; a cause that is extremely fragile can occur only once.
However, to the extent that we are looking (as we are presently) do for The Cause and we want a cause to be The Cause on grounds of its being the closest to being a sufficient cause, then we want The Cause to be a cause of the type-effect. The Cause is a cause of the effect as specified in a non-fragile way. If this hope can be fulfilled, then this proves that it is wrong to think that it is always the case that the more fragile a cause is, the more fragile the effect is. Two examples may clarify this:
I publicly announce that I am going to commit suicide. A pharmacist gives me a prescription for a lethal drug and a friend offers me his house to be left undisturbed when taking the drug. In this situation, my decision to commit suicide is most probably The Cause of the effect, since it makes the greatest probabilistic difference to it (it is, moreover, a necessary condition of it); it is closer to being a sufficient cause (at least under the assumption that I do not often take such a decision). The pharmacist’s prescribing me a lethal drug is a cause of my suicide, but it is not The Cause. However, if we describe what the pharmacist did by saying "The pharmacist prescribed me a lethal drug in the situation in which I had decided to commit suicide and I had made arrangements with someone to let me undisturbed", then what the pharmacist did appears to be closer to being The Cause. In the situation in which I had often desired to commit suicide and did not complete only because of lack of a cooperating pharmacist, the pharmacist’s prescribing me the drug has even higher chances to being The Cause - - because it would be more unlikely given the other causes of my suicide.
However, in this example more detailed specification of the cause means a relational specification: the event is specified (and becomes more fragile) since it is individuated by relation to the other events. If we stick to a non-relational specification of a cause, its turning into a sufficient condition for the effect does no longer look so promising. Think of the classical example with the forest fire that appears because of a lighted match near it. The presence of the oxygen in the air and the absence of a wind towards the opposite direction of the forest are causes of the fire, yet the presence of the lighted match is The Cause. If we non-relationally specify the event oxygen was present there by indicating (if possible) the exact amount of oxygen in the air and the number of molecules and atoms of oxygen present there, it does not seem that the presence of the oxygen should become, by this, a sufficient cause of the forest fire; even less so a sufficient cause of a repeatable (non-fragile) forest-fire This proves that the non-relational specification of a cause may indeed help in making it a sufficient cause of the effect; but a relational specification helps even more. If our task is to turn a cause into The Cause, the advantages brought about by very detailed non-relational specification cannot override the advantages provided by some detailed relational specification.
2) Principle of Probabilistic Difference: The Cause is that among the several causes of an event that makes the greatest probabilistic difference to the causal history of the effect. An alternative naming for this would be The Principle of the Greatest Difference. What this principle says is: when E is an effect of several causes and a ranges among the causes of E, C is The Cause of E if
(For any a)[p(E/C)-p(E/~C)] > [p(E/a)-p(E/~a) (1)
Formula (1) favours necessary over non-necessary causes for the title of The Cause. Indeed, a necessary cause is a sine qua non condition for the effect’s occurrence, and, as such, it is not preemptable: it is not the case that, had a necessary cause not happened, another event would have supplied its causal role. Indeed, for necessary, sine qua non, non-preemptable causes,
p(E/c)=1 and p (E/~c)=0
and
p(E/c)-p(E/~c) = 1, if determinism is true.
Among the causes of an event, some may be fully preemptable, i.e. they are such that, had they not occurred, another event would have been occurred and fully compensated, so to speak, the loss of causal power brought about by their absence. For such fully preemptable causes,
p(E/c)-p(E/~c) = 0 , if determinism is true, since
p(E/c)= p(E/~c)=1
For necessary causes, the difference looked for will be 1-0=1. For non-necessary causes (and I include here both causes such that , had they not occurred, another event would certainly caused the effect - call them fully preemptable causes - and causes such that had they not occurred, another event could have caused their effect - call them potentially preemptable causes) - the difference will be less than 1, since
P(E/c & other causes) =1
And
P(E/~c & other causes) >0.
The last formula captures the intuition that the effect would have had a chance of occurring even in the absence of cause c. In other words, c could have been replaced.
If c were a fully preemptable cause, then it would have been preempted and thus p(E/~c & other causes)=1. Fully preemptable causes make no probabilistic difference to the causal history of the effect, that’s why any of them has the worst credentials, so to speak, to be The Cause.
However, formula (1) may favour some non-necessary causes over necessary causes. If p(E/C) is very small and p (E/C)=0, C is a necessary condition of E. However, there may be another cause D such that p(E/D) is very high and p(E/~D) is reasonably low (although non-null); thus, although p(E/~D)>0, still
p(E/D)-p(E/~D) > p(E/C)-p(E/~C).
Thus, the principle of making the greatest probabilistic difference serves indeed to rule out fully preemptable causes; but it can rule out necessary causes as well. This last consequence is harmless, insofar as it is obvious that an event has many necessary yet trivial causes, including the Big Bang, which we nevertheless do not want to consider as good candidates for The Cause. (It is obvious that any event has many necessary trivial causes at least if we want to keep a reasonable standard of fragility: a staunch opponent of fragility could say that any cause is such that, if it would have been absent, then the same event could - or would - have occurred in a different guise).
The most important consequence of the principle of probabilistic difference is that The Cause is that among the causes which is the least vulnerable to preemption. The Principle of Probabilistic Difference also offers an understanding of the concept of vulnerability to preemption that takes into account not the number of preempted potential causes of an effect, but the likelihood that such preempted potential causes could have been effective. That is, it makes the concept "vulnerable to preemption" closer to "would, almost certainly, have been caused by some other causes" rather than to "could have been, perhaps, caused by many other causes" - which is correct, since, if indeterminism is true, then admittedly any event could have a very slight chance of causing any other event. (Another reason why I understand the concept "vulnerable to preemption" the way I do is that, if I understood it as "could have, perhaps, been caused by many other causes", then disjunctive events would appear as very vulnerable to preemption and disadvantaged ab initio from the contest for being The Cause. I do not, however, want to exclude disjunctive events from the title of The Cause for reasons that will appear below). The Principle of Sufficiency told that The Cause is a cause that is difficult to prevent from reaching the effect; the Principle of Probabilistic Difference tells that The Cause is a cause that is difficult to preempt.
The Principle of Probabilistic Difference vindicates a moral intuition according to which an agent who performs an act whose consequences would have, most likely, have occurred even in the absence of his action is to a less extent responsible for that action than an agent whose intervention was closer to being a sine qua non condition of the event`s occurrence. Thus, it vindicates a common argument that some criminals bring to justify their deeds: ‘It is true that I stole, but if I hadn’t stolen, then someone else would have stolen, therefore I am not guilty’. I believe that common sense partly accepts and partly does not accept this argument: to the extent that it does not accept it, a reason is that we can never be sure about the criminal’s claim that his action was fully preemptable, i.e. it was certain that someone else would have stolen in his stead. To the extent that common sense admits it, the conclusion is that such a criminal is less guilty than a criminal who committed acts which, most probably, noone else in his stead would have committed. For example, in a country plagued by corruption, where many people steal many things, the above excuse for stealing has a grain of truth: it may point to the imperative that preventing corruption is a higher priority than punishing any individual theft.
The Principle of Sufficiency and the Principle of Probabilistic Difference are independent of each other. The Principle of Sufficiency favours causes that are difficult to prevent from causing the effect; the Principle of Probabilistic Difference favours causes that are difficult to preempt and, among these causes, those that are difficult to prevent from causing the effect. It is important to note that the Principle of the Probabilistic Difference thus does not ask The Cause to be a necessary cause: it favours necessary causes since p(effect/lack of a necessary cause)=0. But it requires that The Cause should significantly raise the probability of the effect, i.e. that it should be close to being a sufficient cause. The Principle of Probabilistic Difference, as I formulated it, is biased towards Sufficiency - if it were meant to select only necessary causes, it would be formulated as
(for any cause a of E)p(E/~C)<(p(E/~a)
and it would be a measure of the closeness of a cause to being a Necessary Cause.
The Principle of Probabilistic Difference is however not equivalent with the Principle of sufficiency: this can be easily realized, since an event C may qualify as The Cause under the Principle of Probabilistic Difference if
(for any a)P(E/C) – p(E/~C) > p(E/a)-p(E/~a)
yet
(for any a) P(E/a) < p(E/D)
- where a ranges over causes of E
However, if we test the probabilistic impact of a cause given the other causes of the effect, we reach the unwelcome conclusion that, for any causes c and d (including The Cause),
P(Effect/c & the other causes besides c) = P(Effect/d& the other causes besides d)
since d belongs to the other causes besides c and c belongs to the other causes besides d.
3. Principle of Unlikelihood: The Cause is that among the causes of an effect which is the least probable, given the other causes of the effect In support of this principle, consider the example of a person who is sad (or angry) after playing at the lottery and not gaining. Since not gaining at the lottery was the expected outcome of his participation, I think there is a strong intuition in support of the denial that his not gaining at the lottery is The Cause of his being sad (or angry). (It can be safely assumed that it is a cause, since it raised the probability of his being sad –or angry - compared with the situations in which he does gain at the lottery). But in such a situation, we are inclined to presume that the person was sad – or angry – because of other events or character traits. On the other hand, consider the example of someone who displays the same sadness – or anger – after having loaned a big amount of money and not receiving it back. In such a situation, I believe there is a very strong push towards considering the event consisting of his not receiving the money from the borrower as The Cause of his sadness (or anger). The rationale behind this is that not receiving the money back from the borrower is a highly unexpected event, while not gaining at the lottery was the expected event.
The Principle of Unlikelihood may be misunderstood as saying that people who are consistently and predictably evil are less guilty than those who do evil things unexpectedly. However, my Principle of Unlikelihood says that The Cause should be unlikely given the other causes of the event. That is, The Cause, to the extent at which it is unlikely, is such that the other causes of the event do not usually lead to the effect, and they do not lead to the effect precisely because of the absence of The Cause. The Principle of Unlikelihood does not say that The Cause should be unlikely to occur given its causal history, i.e. that the causal history of The Cause should be, so to speak, one of the most indeterminist regions of the world.
Moreover, the Principle of Unlikelihood is consistent with the Principle of Sufficiency since the causes that occur unexpectedly are more difficult to prevent from reaching the effect: they have a higher chance of being sufficient causes than causes that occur predictably.
This principle allows us to solve another difficulty. On both Lewis’ early analysis of causation as a chain of counterfactual dependence and on Kvart’s analysis of causation as ex post facto probability raising, any cause is effective via a chain of intermediary events, that are themselves causes of the effect. (On Lewis’ analysis, these events are those that depend counterfactually on prior events that in turn depend on the cause; on Kvart’s analysis these events are the channellers, the events that happenned between the time of the cause and the time of the effect and which increased the probability of the effect). However, it is obvious that we must eliminate some of these intermediary events as candidates for the position of The Cause. The two principles presented above - The Principle of Sufficienct and The Principle of Probabilistic Difference - cannot do the job in a satisfactory way, since it may be the case that the last intermediary event would qualify in most cases as The Cause: indeed, the last intermediary event is in many cases the closest of being a sufficient condition for the effect, i.e.
P(E/Last intermediary cause)>p(E/c)
and is very close to being a necessary cause: if such an event were absent, then it would be unlikely that another event compensates this causal loss.
(The last conclusion is supported by the intuition that most preempted potential could-have-been causes distribute uniformly in time; therefore at the time of the last intermediary cause, most preempted potential could-have-been causes are already neutralized; they can no longer threaten to preempt those causes which are causes). The corollary of this intuition is that, in general, there are less candidates for late preemption than for early preemption.
A trivial example is that of a bullet which kills a person. In its trajectory from the gun to the victim’s body, the bullet passes through several intermediary positions. The event E that consists in the bullet’s being one millimeter away from the victim has high chances to qualify for the title of The Cause of the victim’s death according to the Principle of Sufficiency (since it raises the probability of the victim’s death more than many other causes, such as the bullet’s being one meter away from the victim’s body) and would almost certainly qualify for being The Cause according to the Principle of Probabilistic Difference – since it is very difficult to preempt given the short time that remains until the effect. However, in this example event E was very likely given the other causes of the event (first of all, given the fact that the assassin triggered the gun, which we consider to be The Cause of the victim’s death) and it satisfies in many cases the principle of the greatest probabilistic difference, since it is likely to be the least vulnerable to preemption (at least in the cases of early preemption, it is likely that the causal chain leading from the preempted potential cause to the effect was already blocked). Moreover, the non-occurrence of the last intermediary cause would be much more difficult to repair in order to produce the effect at the same time.
However, it would be unintuitive to make the last intermediary causes as The Causes of the effects. An easy way to solve this difficulty would be to say that The Cause should be the beginning of the causal chain; should be the first ring of the causal chain, i.e. it should not itself be caused by another cause of the effect. However, this will not do. The gunseller’s selling a weapon to the would-be assassin is obviously a cause of the assassin’s murdering the victim. But we want, in general, the assassin’s act to be The Cause of the Victim’s death, and not the gunseller’s selling a weapon to him - exception are cases where the assassin was known to be a psychopath who for all his life wanted to murder The Victim and the gunseller, knowing this, sold him a gun.
There is also the trivial objection that the Big Bang or – if you want – God’s creating the world, is a cause of every event. The intuition that The Cause should be the beginning of a new causal chain, should initiate a causal chain of its own is thus wrong and vague - there are several causal chains leading to an event.A better proposal is to say that The Cause should initiate the less likely causal chain to the event. I claim that the principle The Cause is the least probable among the causes of an event captures this intuition.
The principle that The Cause is the least likely among the causes is supported by the following remark about memory: Suppose that you witnessed a certain event. Afterwards, someone whom you trust but who seldom speaks to you tells you that that event really happened. Now you believe that that event happened and perhaps even have a vivid mental image of it; this is however not sufficient for your remembering the event, if we consider remembering as involving a causal connexion between the event remembered and the event of remembering it. If we differentiate between a cause and The Cause, however, we are able to provide a more refined account of such situations: perhaps your witnessing the event left some traces in the present remembering: according to the account of memory one prefers, such a trace may be thought as some traits of the mental image of the event witnessed or some (verbally expressed) beliefs about that event. However, the remembering (which takes place in the present) is not exclusively characterized by these traces of the past event: the fact that someone told you that the event you are remembering really happened alters your present remembering. In such a case (call it quasi-remembering?) I am inclined to say that The Cause of the present event of believing/representing the occurrence of the past event is not your having witnessed it, but your having heard the account afterwards – since it was less likely to happen.
Your having witnessed the event may still count as a cause of the present act of quasi-remembering. On the contrary, if someone often tells you that event E happened and you actually witnessed it, then The Cause of your remembering E is your having witnessed it. This is supported by a folk-psychology intuition about memory, namely that those events which are the least expected are those that leave the most persistent memory traces. The above example shows a way how to deal with at least some cases of trumping preemption: on my account, the preempted potential would-have-been cause may still count as a cause, while the preempting cause is The Cause of it.
The principle "The Cause is the least likely among the causes" may also shed light on the voting paradox, which is a case of multiple preemption. An analysis that takes into account the difference between a cause and The Cause will yield the result that the voter who voted by surprise, who was expected to cast his vote for the would-be-losing party but decided in the last moment, and after difficult deliberations, to vote for the would-be-winner party is a good candidate for being The Cause of the victory of the winner party. (Another intuition that needs to be taken into account here is that the voter who convinced the highest number of previously undecided voters, or of voters decided to vote for a different party, is The Cause of the effect: this intuition suggests another formula, namely that The Cause is that among the causes which is a cause for more causes; however, I think that the principle of probabilistic difference captures this intuition).
The principle that The Cause is that among the causes which is the least probable, given the other causes of the effect, captures an important intuition: we are not responsible for all the (bad) events that happen, inter alia, because of our omissions. We are responsible for those effects of our actions that happen because of our unpredictable omissions. This is the reason why Bennett’s analysis of killing versus letting die was attacked on the ground that it does not distinguish between letting happen and not preventing from happening. According to Fitzgerald (Analysis, 118, 1967, p. 134) there is a difference between not preventing something from happening and refraining from stopping it, the first being expectable and the second unexpected: X refrains from doing something if we should normally EXPECT X to do it (p. 134)
An example given by Hart and Honore may be quoted in this context: A hits B who falls to the ground stunned and bruised by the blow; at that moment, a tree crashes to the ground an kills B. A has certainly caused B’s bruises but not his death; for though the fall of the tree was, [like the evening breeze in our earlier example], independent of and subsequent to the initiating action, it would be differentiated from the breeze in any description in causal terms of the connexion of B’s death with A’s action . (p. 73) – Hart and Honore add that even if A intended his blow to kill B, he should be convicted for attempt to kill with a fatal upshot, but not for manslaughter. A did not cause B’s death even if A intended it.
This example is classified by Hart and Honore as a coincidence, the falling of the tree is considered as the cause of B’s death. (Note that Hart and Honore do not distinguish between causes and The Cause). What is definitory for the occurrence of two events being a coincidence is that the conjunction of two or more events in certain spatial or temporal relations be very unlikely by ordinary standards (p. 74). On Hart and Honore’s example, A’s hitting B is not considered even as a cause of B’s death. This is certainly a wrong conclusion, since A’s hitting appears as a cause of B’s death on a counterfactual analysis (if A hadn’t hit B, then B wouldn’t have died). On the probabilistic analysis of Kvart, A’s hitting B certainly raises ab initio the probability of B’s death and, moreover it raises it ex post facto since the falling of the tree on B is caused by A’s hitting B, therefore it appears as a channeler for A’s hitting B and B’s death.
In this example, A’s hitting B cannot, however, be considered as The Cause of B’s death because (and insofar as) the tree’s falling was unlikely. Nevertheless, this judgment should be altered, in my opinion, if A hit B rightly expecting the tree to fall on B after B’s falling down. Even more: If the tree’s falling was a likely event - I mean, more likely than A's hitting B -, we should still keep A as responsible for B’s death in a strong sense (manslaughter), even if A did not actually expect the tree to fall; we would rightly say "A should have known that the tree was likely to fall on those circumstances". A’s only excuse in Hart and Honore’s example is provided by the low probability that the tree should fall. If, on the contrary, the tree’s falling was a likely event, then A used (deliberately or not) the tree as an instrument (a weapon) to kill B. A would then be guilty, since his act is The Cause of B’s death – although the tree’s falling would still count as a cause of B’s death.
But I also imagine the following example: A shot a bullet to B; the bullet would have hit and killed B instantly, but, before hitting B, a tree fell in the bullet’s trajectory over B, gravely wounding him. The tree’s fall prevented B from being hit, although – as a result of his wounds – B died several days later. In this example, I dare to say that A is guilty of B’s death, since the falling of the tree delayed B’s death. That A caused B’s death is a consequence of another intuition that I am going to analyze, namely that The Cause is that among the causes which accelerates (the most) the effect. (We may imagine that the tree was caused to fall by someone who knew what A did and wanted to save B’s life; and if we assume that B’s death is a morally bad thing under all circumstances, then the person who caused the tree to fall was admirable and successful in his purpose of delaying B’s death).
4. The Principle of Accelleration: The Cause is that among the causes which accelerates (the most) the occurrence of the effect
The last example above suggests that a further intuition is at work when we isolate The Cause from among the several causes of the event. This intuition is that The Cause accelerates the occurrence of the effect. Call it The Principle of Acceleration:
The Cause is that among the causes which causes the effect to happen earlier than it would have happened, had it occurred without it.
This principle serves to discriminate among several causes that are each of them non-necessary. It is related to the intuition that an event which delays the occurrence of another event cannot be a cause of it. However, I am not ready to accept this intuition: the Principle of Acceleration only partially does justice to it. Here is why:
Call a delayer any event D such that its occurrence causes event E to occur later than it would have occurred in its absence. A potential delayer is an event that would or could have caused the effect to occur later than it actually occurred.
Call an accelerator an event A such that its occurence causes event E to occur earlier than it would have occurred in its absence. A potential accelerator is an event that would or could have caused the effect to occur earlier than it actually occurred.
Delayers and accelerators are causes; potential delayers and potential accelerators are preempted potential would-have-been causes. Potential delayers and potential accelerators may be actual events, although they may not be causes. Both delayers and accelerators are to be understood to work in the context of (together with) the other causes of the event: their temporal impact on the effect is tested given the other causes of the event. Both accelerators and delayers are vulnerable to preemption and thus are not necessary causes, nor do they make a great probabilistic difference to the effect. In fact, they are fully preemptable, extremely vulnerable to preemption, causes. They do not cause the type-effect, but temporal versions of them: they fragilely cause the effect. If among the causes of an event there is an accelerator or a delayer, then the accelerator (or delayer) is effective in causing the effect to occur earlier (or later) than it would have occurred in its absence.
What the Principle of Accelleration says is that accelerators are closer to being The Cause than non-accelerators and definitely than delayers. (To further refine by a metric of acceleration: That accelerator which causes the effect to occur earlier than any other accelerator did is The Cause, according to the Principle of Acceleration). What this means is that The Cause - understood now as an accellerator- is that among the causes which makes the greatest temporal difference to the effect. That is:
(Time of E given C and other causes - Time of E given ~C and other causes) >
> (Time of E given a and other causes - Time of E given ~a and the other causes).
It is important to understand what the Principle of Accelleration says and what it does not say. The Principle of Acceleration does not say that delayers cannot be causes of an event. It only says that delayers are disadvantaged in the contest for the title of The Cause. The Principle of Accelleration does not say that there is any important link between being a cause and accelerating the effect. Accellerators and delayers can be accepted as causes of the effect, although they cause only temporal versions of it: they cause it fragilely, so to speak. Therefore, the Principle of Acceleration should not be confused with the claim that delayers cannot be causes.
If the claim that delayers cannot be causes were accepted, then this would automatically exclude late preemption. In the situations of late preemption, the preempted potential would-have-been cause of the effect is such that it would have caused the effect to occur later than it did in fact occurred: it is a potential delayer. In the absence of the preemptive cause, the effect would have occurred later, since the preempted potential cause was a potential delayer. However, this would be too simple a way to eliminate late preemption:
whenever a candidate-to-cause A would have caused the effect to occur earlier than another candidate-to-cause B, we would be authorized to rule a priori that A is a cause of the effect, and not B.
If the claim that delayers cannot be causes were true, this would too easily eliminate late preemption. More importantly, it could make unintelligible cases of early preemption. Situations of early preemption may be precisely such that a delayer, instead of being preempted (like in late preemption) is effective. That is, in early preemption the preempting cause may be a delayer: had it not been effective, the preempted potential cause would have been effective and the effect would have happened earlier. This would oblige us to exclude the preemptive cause from the causes. Therefore: if the claim "no delayer can be a cause" were true, then cases of potential late preemption (i.e., cases where there are two candidates-to-being-a-cause such that one of them would have caused the effect to occur at a different time than the other) would be automatically decided in favor of the potential accelerator; and some cases of
early preemption would receive the counter-intuitive verdict that the preempted potential cause is actually a cause, rather than the preemptive cause. This would happen whenever the preemptive cause would have caused the effect to occur later than the preempted potential cause would have.
There is something acceptable behind the intuition that delayers cannot be causes. However, the thesis put in this form is unacceptable, since it dissolves too easily the problem of preemption. The Principle of Acceleration captures, in my opinion, the grain of truth behind the wrong claim that delayers cannot be causes. According to the Principle of Acceleration, delayers should be excluded from the position of The Cause. But delayers are acceptable as causes. This conclusion is supported by the Principle of the Greatest Probabilistic Difference, which makes vulnerability to preemption an obstacle for gaining the title of The Cause. But the Principle of the Greatest Probabilistic Difference equally disadvantages accelerators from being The Cause.
Delayers can be causes, although a delayer cannot be The Cause. This has another upshot: The Cause non-fragilely causes events to happen. A cause causes it fragilely. I reject the distinction between fragile and non-fragile events and replace it with the distinction between causing in a fragile versus non-fragile way.
Since a cause that merely accelerates the effect’s occurrence is fully preemptable (had it not occurred, the event would certainly have occurred, only a bit later), the Principle of Accelleration seems to be a secondary principle and valid only when all the previous principles cannot help us much in isolating The Cause. What the Principle of Accelleration says is that in situations where several causes of an effect compete for being The Cause, the one who accelerates (the most) the effect's occurrence has more credentials for being The Cause.
This has implications for the temporal specifications of the Thirsty-Traveler puzzle. It has been suggested that any temporal difference between the effectiveness of the actions of the two enemies entails that only the last enemy’s action is a cause of the Victim’s death. This suggestion is consistent with some intuitions about fragility: when two events collaborated, so to speak, in causing an effect and one of them, left alone, would have caused the effect’s being very different (and temporally distant is a species of different) from how it actually is, it is not a cause of it. The assumption here at work is that an event which, left alone, would have caused the effect to occur under a very distant guise from how it actually occurred cannot be a
cause.
This is not true, according to the Principle of Acceleration. In my opinion, if the second enemy (who emptied the victim's canteen) caused the victim`s death to occur much later than it would have occurred in its absence (but given the other causes), it is a delayer and cannot be The Cause according to the Principle of Acceleration. The more a cause delays the effect`s occurrence, the less chances it has to be The Cause. That is, the more the second enemy`s act delayed the victim`s death, the less guilty he is. Likewise: the fist enemy`s act has the higher chances of being The Cause (and the first enemy has the higher chances of being guilty for the victim`s death), the more the second enemy delays the death. In this example, the first enemy`s act (pouring a quickly-acting poison in the victim`s canteen) is not an accelerator: it is only a potential accelerator.
Therefore, in the Quickly-Effective-Poison version of the Thirsty Traveler puzzle, my thesis is that the second enemy (who emptied the canteen) is not The Cause of the Victim’s death. This thesis is very well supported by the intuition that the second enemy delayed and thus prevented the Victim’s death. (Since we are all mortal, preventing a death is a case of delaying it. I am not sure whether all cases of delaying a death are cases of preventing it, but I incline to think so.Indeed, we may imagine the following dialogue:
The Judge: You emptied the victim`s canteen and - - since the victim died of thirst - you are guilty for his death. What you did is The Cause of the Victim`s death.
The Canteen-Emptier: If I hadn`t emptied the canteen, the victim would have died much earlier -
almost surely s/he would have drunk from the poison the first enemy poured into the canteen. So I delayed his death
The Judge: You did delay his death, but we are not talking here about delaying a death. The death happened because of you, and not because of the first enemy`s act. You delayed the victim`s death, but did not prevent it.
The Canteen-Emptier: I did prevent the death, since I delayed it. There is no difference between delaying a death and preventing it, if preventing a death is possible at all. You think that there is a difference between delaying a death and preventing it. If this were true, then anything that could be done to prevent a death would consist in causing it never to happen, or at least making it more likely that it will never happen. I simply do not know what I should do if I wanted to prevent a death, since noone succeeded in this. Perhaps you assume that delaying a death as much as possible is the best one could do in order to absolutely preventing death. But I don’t see why this is so. Asking me to prevent a death is like asking me to become the dictator of the whole world-population: noone has ever succeeded in this, although - like with preventing death - I think most people desired it. It may be assumed that conquering some countries, and then trying to conquer other countries one after another, is the best way to become the dictator of the whole world. But, truly speaking, we cannot know this, precisely because noone has ever become the dictator of the whole world. Perhaps there are other stragegies in order to achieve this task and we don`t know them. Likewise, if the task is prevent that death absolutely, i.e. in a different sense than "causing it to happen later rather than earlier" then I don’t know what to do: it does not go without saying that delaying the death as much as possible is the best strategy. For it is possible that when delaying a death I actually make it more certain that it should happen at some distant point in the future and I actually decrease the chances of immortality.
If we do not accept that the second enemy’s action is less of a cause than the first enemy’s, then we may find ourselves embarrassed if confronted with the second enemy’s claim that his action was in fact saving the Victim’s life, that he lengthened her life and thus, far from being guilty, deserves a reward. (His case would be strengthened if he added: "By preventing the quickly-acting poison from being drunken by the victim, I offered the Victim the opportunity to find a source of water – like an oasis or a benevolent traveler").
I think delayers can be thus kept as causes of effects, although a delayer cannot be The Cause. However, we might be faced with situations where a delayer is close to being a sufficient cause of the effect, perhaps the closest to a sufficient cause. In such situations, a delayer should be considered The Cause, according to the Principle of Sufficiency. (We can gradualize: a delayer is the closest to being the sufficient condition of an effect when it alone gives the effect a higher probability than any other cause gives it). In the often-discussed euthanasia examples, this would entail a physician giving the terminally ill patient a drug that raises the probability of the patient’s death to an even higher degree than the patient’s illness raises the death, but it delays it, may be the author of The Cause of the patient’s death according to the Principle of Acceleration but not according to the Principle of Sufficiency. This shows a weakness in one of the most popular arguments against euthanasia: what if something totally unexpected happens, like a very unlikely recovery or the discovery of an effective cure?. This popular anti-euthanasia argument assumes that delaying a death for a long time is a good thing, although delaying a death for a long time is compatible with raising the probability of the death happening at a certain moment; on the other hand, it is possible to accelerate a death in the sense of causing it to happen earlier than it would have happened, although the probability difference to the death may be rather small. Thus, medical dilemmas can arise in situation where one of the options is "not doing anything" when a patient has an illness that, probably, will kill him in a short time if it kills him at all, and administering a drug that prevents such an early death but makes it certain that the patient will die at some time later.
A Specification of the Thirsty Traveler case may serve equally well: The first enemy adds a mildly lethal poison to the victim’s canteen. This poison is not lethal in all cases (but only in one third of the cases, i.e. it gives a 33% probability to the drinker's death), although it is quickly lethal when it is lethal. The second enemy spills the content of the Victim’s canteen and adds a new, more lethal but slower-acting poison. (This new poison gives a 90% probability of death to the one who drinks it). I do not know what to say about such cases. I admit the inclination to say that it is impossible that a cause which is slowly acting should raise the probability of an event, comparatively to the one that is quickly acting, but I resist this inclination. It is not true that the more an effect is accelerated by a cause, the higher the conditional probability of it given that cause.
The Principle of Acceleration has however seemingly strange consequences for the fragility of the effects. If a cause is an accelerator, it becomes The Cause of the effect, although it actually did not alter whether the effect occurred or not, but only the time at which it occurred. We would like to say: an accelerator, like a delayer, is not a cause of the effect; it is only a cause of the effect occurring under a certain guise, at a certain time. But such causes (call them causes that work fragilely) do not seem very week placed for being The Cause: they make no probabilistic difference to the effect and so they are ruled out by the Principle of the Greatest Probabilistic Difference. Their causal weight is minimal.
To the extent that this intuition is correct, the Principle of Acceleration is a secondary principle. It serves to discriminate among several causes that are vulnerable to preemption. (There is an obvious relation between preemption and fragility: Causes that are vulnerable to preemption cause an event fragilely. They cause it to happen under a certain guise. This does not mean, of course, that a cause that causes the effect`s occurring under a certain guise is ipso facto vulnerable to preemption: it is logically possible that a cause is necessary (invulnerable to preemption) and that it, moreover, causes the effect to occur under a precise guise. Such a situation would obtain for events that are essentially fragile: events that could not have happened under many different guises. Yet, I have some difficulties imagining such events. Historical events seem perhaps the best examples: the fall of Bastille at the 14 of July 1789 seems to be an event that is essentially individuated by temporal coordinates: it would have been a different event if it had happened at another time. This may be true if the consequences that this event actually had are such that they would have been radically different, had the event happened even one day earlier or later. Nevertheless, I continue to think that even such events have causes of their occurrence anytime or as soon as possible. These causes are the best placed for being
The Cause. I doubt that a cause of the Fall of Bastille happening on the 14th of July rather than on the 10th of June may be considered as The Cause of the effect: my intuition is that such a cause is far from being a sufficient cause, and far from making a great probabilistic difference to the effect. That`s why I do not believe that even an event such as The Fall of Bastille on the 14th of July has consequences that would not have happened if the Fall of Bastille happened earlier. I am inclined to think that this is not a logical, but an empirical impossibility.
In the light of the above analysis provided by the four principles I offerred, The Cause is an ideal concept: The Cause of an event is a cause that is sufficient and necessary; it is, as much as possible, a cause of the type-effect and not of fragile versions of it. If it is not sufficient, it is unlikely given the other causes of the effect; if it is unnecessary, it causes the effect to occur earlier than it would have occurred otherwise.
II. THE ONTOLOGICAL STATUS OF OMMISSIONS
I now want to move to the problem of the causal role of omissions. However, the question whether omissions can be causes already assumes the ontological status of omissions. (I want to avoid saying that "there are such things as omissions" because of the obvious awkwardness of this way of speaking). It is precisely this assumption that I am going to analyze.
The question concerning the moral difference of causation-by-acts versus causation-by-ommissions is different from the question concerning the ontological difference between acts and omissions. If it turns out that omissions do not have a special ontological status, different from acts, then the claim that they have a special moral status becomes unintelligible. But if ommisions are ontologically different from acts, it does not follow that they are morally different. Thus, clarifying the ontology of ommissions is the necessary, but not sufficient, condition for claiming their moral difference from acts.
It would of course be wrong to define an omission as a state of an agent that is identical, under all physical descriptions, with the preceding states. Such a definition would be intelligible only to the extent at which one may reasonably separate the internal from the relational states of a person: it does not go without saying that someone who simply "sits down and does nothing" omits rather than acts, if his sitting and doing nothing enters at some moment of his sitting and doing nothing into the causal history of any event: perhaps the refusal to do something that could have prevented or accelerated the effect.
I believe that everything that I wrote above about causes and The Cause can be said about omissions and The Ommission, if by "The Ommission" we mean that among the causes (including omissions) which is The Cause. (I am not interested in clarifying a concept of The Ommission as that omission which is more-of-a-cause if compared to other omissions). The reason why I believe this is that omissions can be causes, just like any other event. However, this view can be rejected on the grounds that omissions are a special kind of event or even that they are not events at all. The next section of my paper is devoted to analyzing claims that ommissions are different events from ordinary causes. My purpose is to defend the admission of ommissions as causes, with the same title as acts.
An obvious objection to granting the omissions the status of causes is that omissions are disjunctive events. When we say that that Fred did not do anything to save the life of the person who was drowning, we are not committed to saying anything about what precisely Fred did: he did anything that is incompatible (contrary) with saving that person’s life. The disjunction of all the possible incompatible things he could have done is the contradictory of saving that person’s life.
This is Lewis's argument against granting omissions the title of causes: according to him, omissions are disjunctive events and disjunctive events do not deserve a place in the causal space. Thus, a Lewisian solution to the problem "what are we to do with ommissions" could be
(a) either to find a way to characterize omissions otherwise than as disjunctive events – if we want them to be causes
With respect to the accusation that omissions are disjunctive events, I dare to say that any event that is not specified in its minutest details is a disjunctive event. (This does mean that type-events are more disjunctive than token-events; for we can identify a token-event, for example, by means of a definite description such as "that unique event such that....", making it clear that it is a token-event and yet leaving it unspecified). Since any event that is not specified up to its minutest details is a disjunctive event, it would be nonsensical to eliminate disjunctive events from the causal space. (Of course, a reason is that elimination of disjunctive events entails elimination of type-events that are causally effective as types; but this is not the only reason since, as we saw, token events may be disjunctive events as well). John’s coming late to dinner caused the host to be upset; we accept this as a genuine causal statement, although we don’t know when precisely John came to the dinner, nor how he was dressed, nor with which foot he stepped first when he entered, nor what his first words were, or whether he spoke at all. "John’s coming late at dinner" is a disjunctive event, yet it is a cause - perhaps The Cause – of the host’s being upset. That is, no matter how he was dressed, or what he spoke, his coming late is a cause of the host’s being upset. What is so special then about omissions? Their simple disjunctive character cannot be enough to rule them as causes. Why is John’s coming late to dinner an action, while Bill’s not taking the precautions against burglary is an omission?
I presume that the reasoning behind those who hold there is a difference between acts and omissions is the following: John’s coming late to dinner is a disjunctive event since its characterization involves disjunctions, such as "he stepped with his right foot first or he stepped with his left foot first", but none of the members of the disjunctions that characterize it are causally relevant when the event is causally relevant. The causal force of this event is concentrated in the non-disjunctive side of its description; its place in the causal space is described by John’s being late at the dinner: the host was not upset because of John’s saying "Hello" on a soft rather than on a loud voice, nor by its wearing dirty, rather than clean, clothes. The host was upset because John was late to the dinner. On the contrary, Bill’s not taking precautions to prevent burglary is a disjunctive event in a different sense: there is here no way to specify the causal power of this event other than by disjunctive statements. Bill was supposed to take precautions against burglary: instead of this, he slept, or drank, or listened to music, or so on. Therefore, omissions are essentially disjunctive events, while ordinary events – that are not specified up to the minutest details – are accidentally disjunctive. The suggestion here is to say that ordinary events are accidentally disjunctive in the sense that our describing of them as disjunctive events is contingent description, in that we could have described them in a non-disjunctive way and allot them the same place in the causal space.
Another way to put this suggestion is the following: events that are accidentally disjunctive (i.e., ordinary events) are such that their non-disjunctive descriptions are rigid designators in the worlds with the same causal history of the effect. John’s coming late to dinner would be John’s coming late to dinner; but events that are essentially disjunctive are such that, if we designate them in non-disjunctive ways, their descriptions are likely to change. Bill’s not taking precautions to avoid burglary actually consists, let’s suppose, in his sleeping during the time when he was supposed to take precautions. But if we describe his sleeping as a cause of the burglary that followed, then the same event we are referring to as his sleeping could have consisted in his drinking, or watching TV, and the causal history of the effect would have remained unchanged.
This objection works against the crude thesis that there is no difference between acts and omissions. I can only answer it by granting that omissions are indeed disjunctive in a different, more essential ways, than ordinary events. This objection does not, however, prevent omissions from being causes. Rather, omissions become multiply-realizable events; they are essentially multiply-realizable. The consequence, that I am ready to assume, is that it is token-omissions that occupy a place in the causal space, if we want to characterize ommissions in a non-disjunctive way; there are no causal laws involving omissions. There may be a nomological dependence between Bill’s sleeping and the burglary in Bill’s house, but there is no nomological dependence between Bill’s not taking the precautions against burglary and the burglary in Bill’s house. Likewise for the situation where the lamp in my room falls on the floor: there is a nomological dependence that underlies "The lamp fell because nothing supported it", but no nomological dependence between "The lamp fell on the floor because I was in the university instead of being home".
I think this consequence should be assumed. The reason is that the only way it may be rejected is arguing for a general causal skepticism about causation. Such a causal skepticism will say: whenever A is a cause of B, we cannot be sure that A was a cause of B qua A, by its being A. A stone weighing 5 kilos was suspended on a rope that was known to support at most 3 kilos; the effect of this event is that the rope broke. The causal skeptic will say: Perhaps it is not the stone’s being too heavy that caused the rope’s breaking, but the stone’s having its particular shape, or its particular color. Kant’s reading Hume is a cause of Kant’s awakening from his dogmatic sleep: the causal skeptic will say: Perhaps it was not Kant’s reading Hume, but rather Kant’s keeping in his hands a book of the weight that Hume’s book had, or Kant’s turning the 5th page from Hume’s book with the force with which he did so, that were responsible for the effect. (The image is made more vivid if Kant’s awakening from the dogmatic sleep is conceived as a neural event…). Davidson's thesis that causation is extensional, while nomological dependences are intensional, is compatible with this causal skepticism. This is unfortunate, in my opinion. I believe that causation is intensional, since otherwise I would have to be open to the possibility that a heavy object's being hanged to a rope caused the rope to break not because the object was heavy, but because the object had the very particular hue of colour it had. (A trope-color: the-color- of-that-object). I grant that, if causation were extensional, then the above distinction between acts and ommissions would lost its point and thus ommissions would be no more and no less disjunctive than any event: we saw that any event that is not specified until its last details is a disjunctive event.
If the above causal skepticism does seem unreasonable, it is not only because we need to keep a place for type causation. Even if we believe only in token-causation, we are still committed, by our causal statements, to saying more than that a certain event caused another event: we want to say that the cause was effective in leading to the event by virtue of its having certain properties, by virtue of its being so-and-so. The fact that the causal skepticism envisaged above is implausible powerfully suggests that causation is not extensional: causation works via a powerful property and should be specified by a description that individuates is as having that property rather than other properties. In the present context, the failure of causal skepticism means that we cannot accept the suggestion that all events are essentially disjunctive and that, therefore, there is a basis of the distinction between acts and omissions. But it does not follow that omissions are not events. Ommissions are events to the extent at which they are multiply realizable. Bill’s not taking the precautions to avoid the burglary consisted, let’s assume, in his sleeping during the time when he should have taken precautions. At other times it consisted, or will consist, in his drinking. And so on. I see no reason why Lewis cannot accept omissions as multiply-realizable events, if he accepts an event such as the poker’s cooling down as a multiply-realizable event. (The poker’s cooling down is realized in this world by the poker’s losing heat; at other worlds it is realized by the poker’s losing caloric fluid). The difference between an event such as the poker’s cooling down and Bill’s not taking precautions to avoid burglary can be one of these three:
1. The poker’s cooling down has a transworld multiple-realizability. But an omission (Bill’s not taking the precautions) is multiply realizable in this world.
2. An ordinary event (the poker’s cooling down) is realized by disjuncts that are not overly distant between themselves. (This is a reason that Lewis explicitly quotes). On the contrary, omissions are instantiated by overly distant disjuntcs. For example – says Lewis – we may accept an event such as talking loudly or talking softly as a genuine event, but we cannot accept an event such as talking or walking as a genuine event. (See Lewis’ article about Disjunctive Events in Philosphical Papers, vol. II).
However, this is a misplaced accusation against disjunctive events, not only against omissions. The trivial reason is that the notion of closeness between the disjuncts is far from clear: are two contradictory, or incompatible, events, closer or less close to each other than two events that share nothing? From Lewis’ example, it seems they are closer. But think of an event such as the plane’s landing, which is, by all standards, a genuine event. It is, as we saw, a disjunctive event in that it means the plane’s landing with or without passengers, the plane's landing with a broken wheel or without any broken wheel and so on. Now, it can be claimed that the disjuncts are close to each other since they are contradictory. Nevertheless, the plane’s landing means also the plane’s landing on a windy or on a calm weather. If this counts again as a genuine event, then there is no reason why the plane’s landing with passengers or on a windy day should not count as a genuine event. But this is a kind of event that Lewis would have rejected, since the disjuncts are here distant from one another. However, any event that is not specified until its minutest details is such an event. Therefore, Lewis’ attempt to use the closeness among disjuncts as a filter for sorting out acceptable events from non-acceptable (because disjunctive) ones is wrong, no matter how we understand the closeness
3. The reason I presented above. Ommissions are essentially disjunctive events, while ordinary events are accidentally disjunctive events. Ordinary events, that is accidentally disjunctive events, are causally effective via the property that expresses the powerful property; while omissions, if they are described in non-disjunctive ways, are described such that it is not apparent how, as what, by virtue of which property, they are causally effective.
This is a good reason why ommissions are different from ordinary events. But it has its limits and I am now going to provide an argument why it has its limits. The reason is that it depends on our language. It depends on an accidental feature of our language: namely that perfect antonyms are very rare. Here the old Wittgensteinian dictum, that we speak about our language when we aim to make ontology, comes true. I can imagine a perfectly intelligible language in which the descriptions for omissions do not come in a negative form. There is no reason why everything we express using negative sentences could not be expressed using affirmative sentences. Some clues exist in our language: to forget means not to be able to remember. Should we consider this as a reason to classify all events of forgetting as omissions? I don`t think so: what if, let`s say, a compact neural basis for all the events of forgetting) is discovered and no compact neural basis for all the events of remembering is discovered? Should we say that it was empirically discovered that forgetting is an act and remembering is an omission? I think it is safe to assume that, if forgetting is a disjunctive event, there is no reason why remembering should not be a disjunctive event as well. As I pointed above, all the events that are not specified up to their minutest details are disjunctive events. Some of these are essentially disjunctive in that their causal power cannot be expressed by an affirmative sentence; others are accidentally disjunctive in that their causal power can be expressed by an affirmative sentence. Ommissions are essentially disjunctive events. But the difference between essentially disjunctive and accidentally disjunctive events is rooted in a contingent feature of our language. Thus, the intuition that a negative sentence is less informative than an affirmative one is true only given a contingent linguistic framework.
I want to stress that I introduced in the picture the problem of the causal skepticism as a difficulty, as a potential counter-example to the thesis that
ommissions are ontologically identical with causes. If we can afford not to care about causal skepticism and say that causation is extensional, then ommissions would a fortiori be causes. This would be a Davidsonian view of causation, according to which "events who enter into causal relations do not care about our description of them". If this view is correct, then whenever an ommission is causally effective, all we can say is that the event that is, in that case, the realization of the ommission is causally effective: it is immaterial to its place in the causal space that we described it as an ommission. (If causation were extensional, then there is no reason, in principle, that we should not describe all the events as ommissions: if we did so, we would not convey any false information about causal relations).
Therefore, my answer to the question "Do omissions exist as genuine events? Can they be causes?" is a double yes. They exist as the bases of the multiply-realizable event that is an omission-as-a-disjunctive event. Describing them as such has the advantage that omissions no longer appear as disjunctive events, but it has the disadvantage that we do not describe them via their powerful intension that ascribes them their role of causes. Ommissions exist, moreover, as omissions, and they can be described via the powerful property that makes them causally effective. This has the inconvenient, however, that their description is disjunctive. But this inconvenient is related to a contingent fact about our language, namely the rarity of perfect antonyms. It is however imaginable that a language translatable into our language should not display this phenomenon. (It is also noticeable that the admission of omissions as causes was done while holding that causation is not extensional. If causation were extensional, then the way we describe the events that are causes would have no relevance and the causal skepticism I presented in the last section would be acceptable. If such a causal skepticism were acceptable, then omissions would be acceptable only as the (token) bases of their disjunctive description. The same fate would befall, however, other disjunctive events: and all events are disjunctive in a way or another.
Since omissions can be causes, then there is no intrinsic moral difference between acts and omissions. This has a clear implication for euthanasia cases: if a doctor allows a patient to die (by passive euthanasia) there is no reason why this act should be morally less objectionable than actively helping the patient to die. And the argument "if omissions are no morally different from causes, then we are guilty for the death of all the people who die by starvation and whom we did not help; we are guilty for all the wars started by dictators whom we did not kill" is a wrong argument. In the euthanasia example, the doctor`s ommission to prolong the patient`s life has high chances of being considered The Cause of the patient`s death: it significantly raises the probability of the effect; it was unlikely, to the extent that the doctor`s intervention to save the patient was the expected thing to do; it is close to being a necessary cause, if the patient did not have a lethal disease. (A clearer example is the classical case of watching someone else drowning and doing nothing to help him: if the person who sits and watches is a good swimmer, his omission to help the drowning-victim was close to being a necessary cause of the victim`s death). On the contrary, in the case of my (alleged) causing by omission a war started by a dictator whom I did not kill (nor did I try to kill), it cannot be said, according to the four principles for determining The Cause, that my omission to kill the dictator was The Cause of the war: first, this omission is by no means a sufficient cause of the war; since it did not significantly raise the probability of the war; second, it is not a necessary cause, since perhaps the situation was such that, if I had indeed tried (or event succeeded) to kill the dictator, someone else would have replaced him and started the same war; third, my omission was the expected event given the other causes of the effect (the war) that had already happened and as such was not unlikely. The claim that we are guilty for the death by starvation of those people to whom we do not donate food is somewhere between the above two examples: my omission was not a sufficient cause of their death (since it is usually not followed by anyone’s death); it was a likely attitude given the other causes of their death; it was not, most probably, a necessary cause, since there is a possibility that even if I had donated food, it would not have reached the intended destination. (Poor countries are usually very corrupt countries, so perhaps the money i would have donated would have been used for maleficient purposes). This does not suggest that we should refrain from helping people who die by starvation, but that eliminating/reversing The Cause of these deaths is likely to be much more laudable than eliminating some causes with a tiny causal impact.
References:
David Lewis – Philosophical Papers vol. II
therefore p(E/C)
(b) or to deny that omissions are causes. If so, we should strive to find a method to find events that could be causes of the effects that are prima facie caused by omissions. It is obvious that denying that omissions can be causes does not mean that events which prima facie happen because of omissions are less caused than events that are caused by actions: if we disqualify ommissions, we need to find better substitutes for their apparent causal impact. I don’t know how such a method might look like, although I will way a few words below about it.
I don’t see why this should be a compelling reason. I see no reason why Lewis, who accepted trans-world-disjunctive events as causes, could not accept events that are disjunctive in the actual world as causes. There are many events that are not ommissions and that are multiply-realizable, i.e. disjunctive, in this world: inviting someone to dinner, calling the police, trying to get attention, getting from London to Paris and so on.
Jonathan Bennett – Killing and Letting Die, in Analysis 1966
Donald Davidson – Causal Relations, in The Nature of Causation, edited by Brand Myles, Urbana : University of Illinois Press, 1976