A Dialectical Discussion of Causation and Counterfactuals

The following discussion will ask the question of whether one can consistently hold a belief in an empiricist epistemology while subscribing to a counterfactual account of causation. The problem comes from the fact that counterfactuals are conditionals where the antecedent of it is a non-existent event, and is thus non-experienceable. It will be argued that it is possible to address this problem if one accepts a contextualist account of causation and a belief that absences are actual events. The discussion will have the following structure. It will begin with a lengthy explication of the relevant aspects of Lewis’ theory of counterfactual causation. Seeing as the above concern affects all variants of Lewis’ theory, no particular version will be discussed in depth, but rather the relevant concerns will be identified. The empiricist challenge will then be explained in more detail before considering two solutions: one is the acceptance that absences are experienceable, and the other is to limit Lewis’ account to deterministic systems. The latter option will be deemed to be seriously flawed, while the former option, when accompanied by contextualist causation approaches, presents a solution. The discussion concludes with reflections on further possible avenues of research.

1. The Lewisian Account

Lewis’ theory of counterfactuals

In order to understand how Lewis understands causation, and its counterfactual nature, one must first understand Lewis’ theory of conditionals. A conditional, traditionally conceived, is a proposition which has the logical form ‘If P then Q’ – that is, if the consequent is true, then the antecedent is true. So, if John is fat then John has been eating too much: John is fat, and so John has been eating too much. This conception of conditional has advantages and disadvantages. The main advantage is that it is intuitive in its pronouncement that if the antecedent is false, then the consequent is false. For example, if I throw this glass, then it will break. However, if I throw the glass and it does not break, then clearly the proposition is false. However, it also carries with it substantial disadvantages in that it offers strange responses where only the antecedent is false. So, suppose a doctor claims that if everyone had leeched their infants last year, then they would grow big and strong. Seeing as no one did leech their infants last year, the proposition is, by the logic of material conditionals, still true. However, it also appears to be clearly false: there is no reason to believe taking the blood out of children facilitates growth (Nolan 2005).

It is attempts to deal with this latter problem of material conditionals that made Lewis (Nolan 2005) create two new forms of conditional: subjunctive (counterfactual) and indicative. An indicative mood is one which purports to state a fact (e.g. If John is tall, then he has gone through puberty). A counterfactual conditional is marked out due to two factors: first, the antecedent is assumed to be contrary to fact (that is, false), second, it is subjunctive in mood (i.e. expresses a condition which is doubtful). There are issues with this demarcation, seeing as linguists have doubted whether the English language has a subjunctive mood different from its indicative. Nevertheless, this discussion will follow Lowe (2002) in taking counterfactual conditionals to generally be of the form: ‘If X had been Y, then Z’, or ‘If John had ate his greens, then he would not be so fat.’ Lewis is not particularly concerned with indicative conditionals, believing a material conditional analysis to be sufficient. Lewis then goes on to discuss what makes a counterfactual conditional true. For Lewis, a counterfactual conditional is true if, in the closest possible world where the antecedent is true, the consequent is true as well. This involves two terms which need to be explicated: possible worlds and their ‘closeness’.

What is a possible world? A possible world is a maximal possible situation. A possible situation is maximal if the following condition is satisfied: given any candidate proposition p, either, in the situation, it is the case that p is true, or it is not true. Possible worlds are mutually incompatible with each other – for instance, considering the above proposition, the only way in which one possible world could be distinguished from another is if some proposition which was true in some possible world was not true in the other. Of course, this also entails that they are not compatible, for if they were compatible, then that distinctive proposition would be true and not true – a contradiction (Lowe 2002). Lewis (1986a) has an idiosyncratic conception of these worlds, as Lewis is a proponent of ‘modal realism’ – that is, the belief that each possible world was a spatio-temporally unified cosmos, of which each possible world was causally independent of the other (Moore 2012). What this, for Lewis, means is that he has an account of the truth of modal propositions. So, for example, if Lewis ever claimed that John would never love Jill, Lewis is saying that in each possible world, there is no (existing) referent satisfying ‘John’ which loved ‘Jill’. For Lewis, also, ‘actuality’ becomes an indexical expression – much like ‘now’, ‘there’, ‘recent’ etc. (Lowe 2002). However, ‘closeness’ is not meant in such an indexical sense (Lewis, when talking about counterfactual conditionals, does not believe some possible worlds are literally spatially proximate to the actual). Rather, what ‘closeness’ means is generally the number of true propositions which are shared between two possible worlds. That is, how ‘similar’ they are. So, if w1 has 50% of its true propositions shared with the actual world, and w2 shares only 5%, then w1 is closer to the actual world. There are issues with this – it is not clear how one allocates weight to certain propositions. For instance, it could be that the 5% of true propositions w2 shares with the actual world are all of the fundamental laws of nature, and there is only a small historical accident which robs it of such similarity, while w1’s fundamental constants could be entirely alien to us, while, again by accident, it happens to resemble the actual world reasonably accurately (Lowe 2002). However, Lowe’s concerns are unwarranted, as Lewis does offer a weighting criteria for ‘similarity’ (Lewis 1986b). The three criteria, in descending order of importance, are (Stalnaker 2015):

1. Fundamentally similar laws of nature
2. Agreement of particular facts (a largely similar past)
3. Lack of ‘small miracles’

This provides some basic criteria with which to distinguish the similarity of worlds. Thus, in the above example, w2 would be more similar to the actual world than w1 because it shares the most important factor: agreement with the laws of nature. Thus, even though only 5% of the true propositions regarding that world are true in our world, these propositions are weighted so much as to provide them with greater weight. Also, even though w1 is superior on the account of criteria ii), it lacks the most important factor and thus is less close to our world. Criteria iii) seems prima facie odd, but it is important for counterfactual analysis. Consider the counterfactual “Had I slept in, I would have failed the exam”. This seems true in conventional cases. However, there are possible worlds where small miracles would enable me to still pass the exam (e.g. issues with the delivery of exam papers, temporary suspension of gravity). Clearly, the counterfactual would not be true if established with reference to that possible world. Nevertheless, one should not use that as the referent for one’s modal claims simply in that, through its use of small miracles, it is less similar to the one where one is punished for a lack of punctuality.

It is now possible to understand Lewis’ theory of counterfactuals. Lewis argued that a counterfactual is true if, in the most similar possible world where the antecedent is true, the consequent is true as well. So, for Lewis, to inquire about the truth of the counterfactual: “If John had ate his greens, then John wouldn’t be fat”, one is meant to consider the closest possible world where John did eat his greens, and inquire as to whether he is fat.

Lewis’ Theory of Causation

It is unfair to criticise a thinker on the basis of what they did not intend to do – to criticise Einstein for failing to contribute to English literature is to err – so it is important to clarify what Lewis’s intentions were in interacting with issues of causation. Lewis was attempting to do conceptual analysis – that is, Lewis was attempting to create a definition or account of causation which was compatible with one’s intuitions, tested against a range of hypothetical cases (Hitchcock 2015). Thus, Lewis’ account of causation must be compared with the above litmus test to consider how well it achieves the above ends. Another important factor to note is that Lewis’ counterfactual account of causation changed throughout his career – with three main periods of qualification and amendment to the original formula (Lewis 1973, 1986c, 2000). However, the criticism raised of Lewis’ account affect all versions of his theory.

A counterfactual account of causation generally is taken to follow the below schema (Nolan 2005, Lowe 2002):

Counterfactual Causation: Event c was a cause of event e IFF c occurred and e occurred, and if c had not occurred, then e would not have occurred.

This introduces the concept of an event – an event is simply some spatiotemporally delineable region in some world (possible or actual). Furthermore, events possess a property called ‘fragility’, which is their ability to be altered through some modal alteration. Thus, if an event would not have occurred unless an extremely specific situation obtained, then the event is very fragile. If an event was likely to happen come what may, then it is not very fragile. Furthermore, in order for c and e to be causally related, the two events have to be logically and mereologically distinct. What this means is that the occurrence of c cannot logically entail e. Suppose Jack violently beats John. Although it is true that, without the beating, the violent beating could not have occurred, it would be wrong to say that the beating ‘causes’ the violent beating in the requisite sense – as they are clearly not two distinct events. Furthermore, regarding the mereological claim, this simply means that the inclusion of a small event within a larger event does not constitute the larger event causing the smaller event – one can call this the ‘non-overlapping’ criterion. Thus, the ‘fractious relationship of John and Jack’ does not cause John’s violent beating, as the beating is simply part of the larger event of the relationship (Hitchcock 2015).

It is important to include this clarification of the above formula, for it prevents immediate, but unfounded, criticism. For instance, Lowe (2002), looking only at the formula, argues that this conception of causation is false, for it leads to the following faulty analysis. Lowe asks us to consider the following counterfactual claim: If Napoleon had not been born, then Napoleon’s death would have failed to occur. This, if one only looks at the above formula, seems to be a convincing counterclaim to the Lewisian analysis, as it strangely seems to imply that Napoleon’s birth caused his death. Lowe defends the claim on the basis that, although counter-intuitive, it may be true – and this argument could work. However, this is beside the point, as the example does not satisfy the logical and mereological conditions above. It does not satisfy the logical condition as, assuming that immortality is impossible, the birth of Napoleon entails his eventual death. Secondly, it fails to satisfy the mereological condition, as both the event of Napoleon’s birth and Napoleon’s death are plausibly part of the same event: Napoleon’s life. Thus, this criticism fails, as it does not respect the requisite qualifications.

Another important aspect of the above formula to take into consideration is that it, for Lewis, does not represent ‘causation’ so much as ‘causal dependence’. The latter is stronger than, and not conceptually equivalent to, causation. This is because causal relations are transitive. That is, if A causes B and B causes C, then A causes C. This relation does not hold for the counterfactual ‘causal dependence’ relation above, as it would be possible for A to fail to occur, and C to occur. For Lewis, the concept of causation is equivalent to a chain of causal dependencies (Nolan 2005). Another qualification is that the counterfactuals cannot be ‘backtracking’. What this means is that one cannot take some counterfactual which states that the past would be different if the present were somehow different, as a genuine causal statement. An example of a backtracking counterfactual would be “If I botched the piano recital, it would have been because I had been drinking the night before”. One of the reasons why Lewis dislikes backtracking counterfactuals so much is because they rest uneasily with ii) above – the idea that a criterion of the closeness of a possible world is that they have similar pasts. A counterfactual which attempts to infer causation from the present to the past would rest uneasily in readily identifying similarity on the basis of such a criterion. Perhaps it would be possible to create a theory of counterfactuals which enjoyed backtracking, but this is beyond the remit of this discussion and it will suffice to say that Lewis himself was not inclined to admit them.

2. The Empiricist Challenge

Lewis proclaims himself to be an epistemological empiricist. For Lewis, what this means is that he believes that the basic evidence attesting to the truth of propositions is perceptual experience and memory (Lewis 1999). This is a fairly conventional position and, although not fleshed out in much detail here, it is not traditionally seen to be incompatible with most doctrines. Now that one has established his epistemology, it is also important to examine, in broad terms, his metaphysical theses – simply because it is fairly uninformative to have a theory of how we know x, without also a corresponding theory of what sort of thing x is. Lewis subscribed to a metaphysical theory he referred to as ‘Humean supervenience’ – the thesis that the world was composed of individual particular facts. The program, in broad strokes, for Lewis, was to provide the truth conditions of all of these individual particular facts (Ismael 2015). Naturally, then, the corresponding empiricist project is to ask the question of how we could come to know these multiple individual facts. This is, again, not particularly controversial and empiricism is hardly in disrepute in an attempt to engage with that fact.

The issue is how it is possible to make such an empiricist method not be in tension with the belief that counterfactual causal claims can be known to be true. Thus, the challenge one sets to Lewis is:

Empiricist Challenge: To what extent can counterfactual causal claims be known to be true under an empiricist epistemology?

One might be perplexed and wonder why this is a challenge – where does the tension between the two lie? The first issue regards the definition of a counterfactual. As stated above, a counterfactual claim is one where the antecedent of the claim is assumed not to have occurred. The challenge which the empiricist is raising is, if knowledge is based on experience and memory, one cannot have an experience (or a memory of experiencing) some event which has ex hypothesi not occurred. The second issue comes with general claims regarding how an empiricist can reconcile oneself with modal claims. Quine, the ‘arch-empiricist’ was famously scathing about de re modality for reasons somewhat related to this fact (Follesdal 1968). But the claim is especially damaging for Lewis and his modal realism. In spite of Kripke’s famous claim that Lewis believed that possible worlds were actual objects observable through a telescope (Noonan 2013), the opposite in fact is the case[1]. As mentioned above, possible worlds are spatially and temporally isolated and suffer from an entire causal disconnect with the actual world. This means that it is ex hypothesi impossible for one to have empirical knowledge of these possible worlds. This does not entail that one cannot have knowledge of these worlds – one can make inferences, and these inferences could in principle be correct, but a committed empiricist, and a committed Humean presumably would not rest easily with that. Inferences are famously flawed epistemological instruments and it is made worse that one cannot check the accuracy of one’s inferences through references to some experience.

There are immediate ways which one perhaps could see in relaxing the above tension. One is to introduce some epistemic humility into Lewis’ account by either denying that modal-cum– counterfactual claims are true, or, by another method: deny that their truth could be known. Lewis himself could hardly endorse either of these options. There are two main reasons for this. The first is that it would entail the failure of his platonic enterprise of providing an accurate conceptual analysis of causation – simply because, conventionally, people accept that causal claims can be true and can be known to be true. An error or non-cognitivist theory of causation is presumably unappealing, and it raises the question of why Lewis spent so much effort writing on an issue which is apparently epistemically bankrupt. The other reason why Lewis would presumably find this move anathema is due to the fact that Lewis also ascribes an incredible amount of importance to causality, arguing that causation is the fundamental element of explanation – that is, Lewis believed that all (good) explanations were causal in that they identified the requisite causes for such an event to occur (Nolan 2005).

Looking into the Void: Can an empiricist experience absence?

Another, somewhat Sartrean (McCulloch 1994), way of addressing the above is to acknowledge that one can experience ‘lack’ in events and that this feature, plus some rational inferences, can lead to causal knowledge in Lewis’ sense. What lack means is that one can experience the absence of some object or event – if this is possible, then one has addressed part of the issue of the above challenge. The issue is that counterfactuals, by their very nature, describe unobservable and absent events. If absent events were tolerable on an empiricist framework, then one is potentially flirting with success. In fact, this move seems to be open to Lewis as well. Lewis provided a conceptual analysis of experience with several criteria, however this discussion will only consider two which might be problematic for the above argument (Stoljar 2015):

• Causal Thesis: When X has an experience, X is in a mental state which typically causes X to act in a certain way, to enter mental states, and is typically caused by things in the world.

This, on some readings, would show that Lewis could not countenance absent events being experiential, in that it has to be caused by something in the world. However, this begs the question against those who wish to admit absent events. If one observes the above definition of an event – a spatiotemporally delineable region of some world – then there is no reason to think that something not occurring in a candidate spatiotemporal region would not constitute an event. For instance, suppose John is excited to interview a candidate for a job interview, and the candidate fails to show up. John is disappointed and this causes him to call the interviewee to see where they are. This seems to satisfy the Causal Thesis, and thus there is no prima facie reason that one cannot experience absent events.

• Ability Thesis: When X has an experience, they can i) imagine ii) remember and iii) recognise the experience

This perhaps would be problematic for the absence theorist, in that it is difficult to literally ‘imagine’ (as in, to construct a mental image of) something which by definition is non-spatial. Attempts to get around this by imagining one’s disappointment fail, as they are identifying an effect of the experience, rather than the experience itself. However, it does not seem implausible to say that one can remember someone not showing up to a meeting – people do remember the tardiness of their peers, and they seem perfectly capable of recognising the event and imagining the somewhat empty room. Thus, there does not seem to be anything in Lewis’ account of experience which precludes his acceptance of absent events.

Lewis (2004), however, simply rejects this course and refuses to admit that absences are events at all. Lewis’ argument for this is that it is the best possible option out of all the other ways of dealing with causation by absence (e.g. John died because he did not eat food). One can grant this argument because the point is simply to establish that Lewis himself cannot admit that absences are events. This does lead to some substantial exegetical issues, as Lewis also acknowledges that events can be causes and effects – a point which directly contradicts the formula provided for Counterfactual Causation above, seeing as it depends on the concept of an event. It is not clear how Lewis can ‘get away’ with such a modification seeing as he is therefore denying that the causal powers of the world are not exhausted by spatio-temporal occurrences (events) – a claim which stands in considerable tension with most empiricist thought. Furthermore, it is also entirely confusing how things which are non spatio-temporal can have causal power at all. How do these ‘non-event’ absences influence other events? That is, why does John die if he does not eat, if ‘not eating’ is not an event? Furthermore, how is ‘not-eating’ not an event, considering the fact that the period of someone not eating is describable in spatio-temporal predicates (Nutrients not entering John’s digestive tract for a duration of t seems like something which is spatio-temporal). In spite of these confusions, this is Lewis’ position.

Nevertheless, the fundamental question is whether Lewis believes that absences are possible to experience. This is not an easy question to answer. Lewis (2004), in the article where he denies absences as events, discusses ‘void’ – that is, something which is in between objects, but is not spacetime, but rather a direct distance relation. Lewis does not assert that ‘voids’, as existent objects, are conceptually impossible – Lewis just does not think they are compatible with the current laws of nature (Lewis 2004). It is somewhat confusing why he asserts this seeing as, in acknowledging that absence can cause events, also means that something not predicable in our spacetime can (and does) exist, and does not violate the laws of nature. Yet, the question is whether absence is possible to experience in light of this. In acknowledging that it is conceptually possible for voids to exist, Lewis acknowledges that some possible world can possess them. If this possible world can also have humans, then, plausibly, humans can experience the void, assuming that humans have such a capacity in that world. Yet, with reference to the criteria of similarity between possible worlds, such a world would be so alien to our actual world that it would be simply to say that a human could never experience a void – seeing as it would involve a different fundamental law of nature, and substantially different cognitive powers on the part of humans. The reason why it is difficult to answer the question properly is that Lewis thinks discussion of voids and absences are similar – but in a considerable number of senses they are not. Voids are by definition non-spatio-temporal, yet absences are spatially and temporally predicable. Voids occur in very far away possible worlds, yet absences are ordinary occurrences. It is not clear why Lewis felt they were important to discuss side by side, or why he felt that one contributed to the discussion of the other. Nevertheless, Lewis’ response here will be taken as a constraint on future theorising: Lewis does not believe that absences are events, and are thus non-spatio-temporal. Following Kant, the following will assume that something being spatio-temporal is a necessary condition for its perception, and so Lewis will be seen as arguing that absences are non-perceptible, and considering his definition of empiricism, thus not within the remit of proper epistemology.

3. Deterministic Systems

Seeing as Lewis finds that absences are not events, and are thus not spatio-temporal (and thus, plausibly, not a possible object of experience), Lewis needs another argument to reconcile his empiricist methodology with his counterfactual approach to causation. One candidate way in which Lewis might be inclined to pursue this is by restricting his analysis of causation to apply only to causally deterministic systems. Although there are several definitions of causal determinism, this discussion will follow Hoefer (2016):

Causal Determinism: Some world is deterministic IFF, considering the state of affairs at time t, any future event from t (t1, t2…) is necessitated by natural law.

Some of these terms require unpacking. A world is as it was above – any maximal possible situation. Natural law just refers to any supervening nomological system which governs the relationship between events – it is nomological in so far as it applies universally to the objects subsumed under the law (e.g. so far as objects share some property which is subsumed under the law, the law will govern the nature of the objects’ relationship with future events). Finally, one must establish the nature of necessity. A conventional way of understanding necessity is that of the material conditional explained above – that is, if some event E happens, the E1 will occur. A stronger version of this relationship would be that of the biconditional: that if and only if E happens, then E1 will occur. Trivially, however, any conditional in a deterministic universe is also a de facto biconditional, seeing as only one set of events could occur. However, one should follow Lewis’ (1986d) definition of necessity, which claims that X is necessary IFF whatever else would be the case, X would remain the case (Williamson 2007)). This defines necessity in terms of the fragility of propositions – necessary propositions are the least fragile propositions possible in Lewis’ schema.

How does limiting the application of Lewis’ counterfactual analysis of causation succeed when limited to deterministic systems? The answer is that, if a system is deterministic, whereby E1 occurs IFF E occurs, then, although one cannot observe the antecedent of a counterfactual, one can infer it. Although this might seem prima facie odd, it is possible to explain with reference to a neurobiological case study. First, one must introduce the concept of a mechanism. Something is a mechanism if it is an organised activity or entity, organised so they are productive of regular changes, beginning from some initial condition, to some terminal point (Machamer, Darden and Craver 2000[2]). Each individual point in the sequence determines the next part. A mechanism is different from the definition of the above in that it is not a deterministic world, but a deterministic system (seeing as systems are not maximally descriptive states of affairs). Roughly, one can characterise a deterministic system (or mechanism), by the following schema: A -> B -> C, with each stage entailing the next. However, the sense of entailment in Machamer et al is not as strong as Lewisian necessity. Lewisian necessity says that X is necessary if X would obtain no matter what else is the case. However, Machamer et al believe a deterministic system is regular in that, under the same conditions, the same entailments would obtain – thus, for Machamer et al, if different conditions obtained, then the system would alter. For example, if one has a computer which performs some activity, Machamer et al would say that the entailments would obtain if all else remained equal – that is, their determinate generalisations depend heavily upon ceteris paribus clauses. However, this is not enough for Lewis; in order for a mechanistic system to have necessity, it would have to still carry out A-> B -> C even if the computer was thrown into the sun.

Let us now consider a case study as to how mechanism (deterministic systems) can provide some assistance for reconciling empiricist accounts with counterfactual causation. Consider the chemical transmission at synapses. This process is that of converting an electrical signal in one neuron, into a chemical signal in the synapse, back into electrical signal in the second neuron. Following Machamer et al’s above schema, one must identify the initial conditions – conceived of as taking some time t at the beginning of the process. One identifies the relevant structural, spatial and, if one is adopting a more nuanced picture, other relevant factors such as the enabling conditions, and places them within the picture of the initial conditions. In the case of chemical transition, one would mention the corkscrew shape of the protein (the voltage gate) and one would also mention the pore lining. Furthermore, one has to address the termination point of the system. Machamer et al take the endpoint to be the conjunctive realisation of two events: the increase in intracellular Na+ concentration as well as a corresponding increase in the voltage of the membrane. Although Machamer et al provide an in-depth account of the stages between these two conditions, there is no real need for the purposes of this discussion – the question is how these two link towards yielding an empirical analysis of counterfactuals.

Consider the following counterfactual:

Had the corkscrew shape of the protein changed, there would be no corresponding increase in the voltage of the membrane.

This, it seems, is true. Yet one cannot observe the corkscrew shape of the protein being different ex hypothesi. However, knowing how the system works, and the related mechanisms between them, enables one to infer the importance of the corkscrew shape for yielding an increase in the membrane’s voltage. Thus, in deterministic systems, where one has a basic understanding of the pattern of entailment, and relationship between them, then one can infer, from observations, the truth of the above counterfactual. That is, when one is presented with a deterministic system, such as a mechanism, one can attest to the truth of the counter-factual claim above, in so far as one is aware that the corkscrew shape must obtain for there to be a corresponding increase in the membrane’s voltage. To simplify the argument: if one comes to know, through empiricist methods, that B occurs IFF A obtains, then one is also justified in knowing that, if A does not obtain, then B will not occur. Thus, assuming that the empiricist is entitled to ‘B occurs IFF A obtains’, then the empiricist is also entitled to subscription to the requisite counterfactual. Thus, even though the antecedent of the counterfactual is ex hypothesi non-existent (and thus cannot be experienced), the empiricist is entitled to believe the truth of the counterfactual. In summary, so long as Lewis restrains himself to determinist systems, the counter-factual conditional will work.

This argument, however, suffers from several fatal flaws. One can concede to the system-determinist that if the empiricist was entitled to ‘B occurs IFF A obtains’, then they would subsequently be entitled to belief in the counter-factual, but Lewis would deny that they were entitled to that claim. It is simply not the case that “Had the corkscrew shape of the protein had been different, there would be no corresponding increase in the voltage of the membrane”, because it is not a necessary truth that the membranes voltage is dependent only on the corkscrew shape of the protein. It could be the case that one could develop some technology which imitates the corkscrew shape of the protein, or perhaps some kind neuronal angel intervenes. The point is simply to observe that the IFF clause is not one which the determinate system theorist is entitled to – and seeing that this is the form of necessity that Lewis himself is interested in, it is unlikely that he would find the move convincing. A possible rejoinder to this counter-argument is to weaken the claim from biconditional to a mere conditional, and include references to how things would be in the closest possible world. But, in that case, it is left confused in what sense this system could even be called ‘deterministic’, seeing as, in order for that argument to be coherent, it would have to be the case that the event A does not entail B – to speculate about the closest possible world where A does not entail B, and where this entailment is meant to be necessary, does not make sense as, if the entailment was necessary, there would be no possible world where A did not entail B (similar concerns about counterfactual analysis underdetermining events can be found in Collins, Hall and Paul 2004).

But the above is not the more problematic part of the system-determinist attempt to solve Empiricist Challenge. There are two other substantial issues. The first is that it is unclear what the counter-factual account of causation is doing in the above scenario. The ability to create a counter-factual claim can only be arrived at after someone has demonstrated that A <-> B. Assuming that A <-> B is interpreted as a relationship of causal necessity, and if the counterfactual claim can only be subscribed to after A <-> B has been established, then the counterfactual formulation of the relationship is entirely parasitic on some other theory of causality. This is because if one interprets A <-> B in counterfactual terms, then one is assuming the truth of the theory in order to prove its truth – that is, it is begging the question. That said, the response does concede that Empiricist Challenge has been addressed. Under the above argument, one can follow an empiricist methodology and come to know the truth of some counterfactual claim. However, the question is what reason there is to care about it, seeing as the counterfactual formulation provides no explanatory weight, and, in its parasitism on the other concept of causation, which established A <-> B, means that the counterfactual account could not be the appropriate analysis of causation. Thus, Lewis has answered Empiricist Challenge, but failed in his larger goal of providing a successful conceptual analysis of causation.

The second substantial issue is that it plausibly fails to abide by Lewis’ stipulation regarding the distinctiveness of events. If one considers the criteria dictating the distinctiveness of events, the events had to be i) logically (the occurrence of c does not entail e), and ii) mereologically distinct (one event should not be part of the other). There are some reasons to suspect that approaching counterfactuals via the above method does not respect these two constraints. Firstly, if it is necessary that A entails B then it seems to be the case that one violates the logical distinctiveness of events. This argument can be accused of conflating different forms of necessity: logical necessity and physical necessity. Generally, causal determinists accept that their determinism is not logically necessary, but is a contingent feature of the world one inhabits – it could have been the case that the laws of nature were different. Perhaps – but this raises some serious concerns about the utility of counterfactual reasoning. Assume two things: that the actual world is deterministic, and that what makes the world deterministic are the laws of nature. Thus, from some time t, every future event was ‘guaranteed’ to happen. This has serious entailments in that any other possible world which diverges from the actual in some candidate counterfactual, must diverge from the actual world in some fundamental law of nature (assuming that the world is deterministic). Considering the great weight applied to the laws of nature in measuring closeness, any other world could plausibly be said to be not close simpliciter. The question then is, for a determinist, there is little utility in using counterfactual as any other possible world is too divergent to yield interesting analysis (For a similar argument, see: Adler 1980). The natural route of this is then to deny a deterministic universe, but this raises problems for those who argue on the basis of determinant systems as a solution to Empiricist Challenge. If one assume that determinant systems could only occur in a determinist universe, then the above attack applies to them. If one argues the contrary –  that determinant systems can occur in non-determinist worlds –  then one has to offer an explanation as to how this is possible. Finally, and briefly, the mereological problem could also attack the determinant systems theorist; just as how a passionate kiss is not distinct from the relationship within which it takes part (Hitchcock 2015), it is an open question how distinct a single process is in the mechanism it is a part of.

4. Context Sensitivity?

Considering the flaws of Lewisian accounts, it is worthwhile to see if it can be modified in an appropriate way so as to strengthen the thesis, and, potentially, make it able to resist Empiricist Challenge. One of the most famous criticisms of Lewisian theory is that of the contextualists (e.g Schaffer 2013, Menzies 2004). Contextualism is the thesis that causation is not some absolute relation specifiable independent of context, and that these contextual elements are necessary in specifying the truth conditions of causal claims (Menzies 2004). Approaches such as this appear to embrace counterfactual analyses of causation but seek to apply stricter limits on relevance of certain counterfactuals (e.g. Hitchcock 2007). First, it would be useful to understand what motivates contextualism and why it is taken to be an improvement of Lewisian theory, before, secondly, seeing whether the improvement potentially offers a solution to Empiricist Challenge.

The main issue which afflicts Lewisian theory from the point of view of the contextualist is the issue of profligate causes. Consider the following example. John is sitting at the pub when he spills a very angry man’s beer. The man, in a rage, attempts to punch John, but instead knocks acid onto a student’s computer, ruining their work. The issue from the contextualist perspective is that the student’s work being destroyed could have been prevented by a plethora of reasons: if the puncher was not quick to anger, if they had elected to work somewhere else, if John had not been born etc. On Lewis’ Counterfactual Causation, all of these elements would count as causes, but, for the contextualist, that is too much – one should be more discerning when one embarks on conceptual analysis. If the student was asked why their work was late, and they responded with “The birth of a stranger named John”, it might be correct, but one would presumably not be impressed by the response. Menzies (2014) believes that accepting the contextualist approach entails a revision of how one embarks upon the analysis of counterfactuals.

While Lewisian analysis is provided as:

Lewisian Counterfactual: If c had not occurred, then e would not have occurred.

The contextualist would prefer:

Contextualist Counterfactual Causation: If c* had occurred rather than c, then e* would have occurred rather than e.

Where the events which are meant to be compared with the actual events are chosen based on their utility and relevance. So, when the student explains why his work was missing, he would make reference to more relevant factors than the birth of John. One does need to provide a theory for discerning relevance, but it is sufficient at this point simply to state that this is the main difference. Can Contextualist Counterfactual Causation provide a solution to Empirical Challenge? No, but it is still superior to Lewis.

Consider the above discussion of whether absences are causal. Schaffer (2005) believes that contrastivism can provide a resolution (contrastivism is a counterfactual formula whereby more than one event is contrasted with another – exactly as in Contextualist Counterfactual Causation, where one contrasts c*/c and e*/e). Schaffer identifies 7 remarks about absent events which are mutually inconsistent:

1. Absent causation is intuitive
2. Absences can explain and predict causes and effects
3. Absences play the same moral and legal role as causes (consider issues such as neglect)
4. Absences mediate causation by disconnection (that is, some events can be explained only in terms of someone enforcing an absence – such as linking the denying of someone food to their death)

On the negative side:

5. Absence causation is counter-intuitive: “The hospital burnt down because Stephen Hawking did not put out the fire” is an odd remark to say.
6. Absence causation is theoretically problematic: it is not clear what kind of event (if any) an absent cause invokes
7. Absences are metaphysically abhorrent – there is no physical process linking the hospital burning down to Stephen Hawking.

Schaffer believes that this ‘paradox’ can be resolved with the adoption of two theses:

1. Adopt negative nominals as referring to actual events
2. Treat ‘absence talk’ as comparing the absent event to the actual event.

What ii) means is to follow the above Contextualist Counterfactual Causation schema, and say, for example: It was the driver’s inattentiveness at the wheel, rather than careful driving, that caused them to crash rather than get home safely. The contrastive strategy does help resolve the above paradox. For the first four points, the contrastive move helps to allocate truth conditions not problematically to absent event talk – seeing as one is referring to actual events, and references to actual events have truth conditions. Schaffer does provide an account for how it deals with 5, 6 and 7 as well, however, the concern of this discussion is not to resolve Schafer’s paradox, but to see whether Schafer’s two theses help us address the Empiricist Challenge.

It seems that they do. As was said above, the question is how to reconcile an empiricist epistemology with knowing the truth of counterfactuals. If one accepts i), that is, that absences are events, and thus experienceable, then there is no real dilemma. Thus, even if the antecedent does not occur, one can still experience its not occurring, and there is no substantial gap between counterfactual accounts of causation and empiricism. This then raises the question of what contrastivism contributes to this debate – after all, if Lewis had, above, accepted that absences were spatio-temporal events, then Lewis presumably could also be entitled to reconciling empiricism and counterfactual approaches. This might be true, but it would have been considerably more metaphysically disorganised. Lewis, in effect, acknowledged when the contextualists charged him with causal profligacy, that they were correct (Lewis would find that ‘the big bang caused my homework being late’ is a true causal utterance) (Menzies 2014). Yet, if Lewis acknowledged that absences were events as well, then his metaphysics would offend those who had a penchant for ontological desert landscapes – this is because every event, and every absence (which are also events) would cause everything.

The acceptance of such a loose notion of causation, although not demonstrably false, is trivial and useless for most purposes. For instance, Lewis believes that all explanations are causal explanations (Nolan 2005), but, if everything that was and was not was a causal event, then it seems difficult to think of how that thesis becomes interesting. Thus, it would be equally legitimate for Lewis to say, in relation to the above neurobiological case study, that the shape of the proteins cause a certain charge of the membrane. Yet Lewis would also find it just as legitimate to provide an infinite list of other causes due to the admittance of all necessary events and absences. Thus, when one considers Machamer et al’s (2000) conception of mechanism, which paints the picture of mechanism as A-> B -> C, there is real no reason for Lewis to accept any delineation there. To specify A as the cause of B is arbitrary for Lewis, as there is no reason to accord priority to A as the cause of B, over some other event (or absence) necessary for B’s occurrence. In effect, Lewis would have to explain the cause of B in terms of the history of the entire world

This is why contrastivism is important. Not many would assent to saying such a metaphysically engorged concept of causation is either their own, or that it is useful. However, the delineated nature of contrastivism helps the picture retain some appeal – just like one wants to know what specific event caused another, one would also want to know how some specific absence caused another. It is difficult to see how Lewis could deliver either of those answers, yet, with the acceptance of i) and ii) above, Lewis could supply a more analytically intriguing, and epistemologically coherent conception of causation.

5. Conclusion

It has been argued here that one can provide a solution to the Empiricist Challenge, which questioned the compatibility of empiricist epistemology with counterfactual accounts of causation, if one accepts two theses. These are that absences are actual events (and are observable) and that one uses a contextualist account of causation. There are some potential avenues for further research. One would be the way in which absences have been treated as observable. This discussion has assumed that if something is spatio-temporal, then it is observable, and that, seeing as events are spatio-temporal, they are observable. However, one might want to deny this inference and claim that not all events are observable, or that events are not spatio-temporal in the right way. It would be important to see how such a position would affect theories in the philosophy of perception. Second, said it has been argued here that a contextualist theory is required for the resolution of the Empiricist Challenge, however it has been silent on which. Further research could inquire into which version of contextualism is best suited for addressing the above concerns, or whether contextualism is viable as a theory in and of itself.

Bibliography

Adler, R. (1980) ‘Possible Worlds Counterfactual Theories of Causation’. Canadian Journal of Philosophy. 10 (Supplementary Volume 6). 119-138.

Collins, J., Hall, N. and Paul, L.A. (2004) ‘Counterfactuals and Causation: History, Problems and Prospects’ in (eds.) Collins, J., Hall, N. and Paul, L.A. Causation and Counterfactuals. London: The MIT Press.

Follesdal, D. (1968) ‘Quine On Modality’ Synthese. 19. 147-157.

Hitchcock, C. (2007) ‘Prevention, Preemption and the Principle of Sufficient Reason’ The Philosophical Review. 116. 495-532.

• (2015) ‘Lewis on Causation’ in (eds.) Loewer, B. and Schaffer, J. A Companion to David Lewis. Oxford: Wiley Blackwell.

Hoefer, C. (2016) ‘Causal Determinism’ in (ed.) Zalta, E.N. The Stanford Encyclopedia of Philosophy. (Available online at: https://plato.stanford.edu/entries/determinism-causal/) [Accessed: 25/04/17]

Ismael, J. (2015) ‘How to be Humean’ in (eds.) Loewer, B. and Schaffer, J. A Companion to David Lewis. Oxford: Wiley Blackwell.

Lewis, D. (1973) ‘Causation’. Journal of Philosophy. 70: 556–7.

•  (1986a) On the Plurality of Worlds. Oxford: Blackwell.
• (1986b) Philosophical Papers Volume II. Oxford: Oxford University Press.
• (1986c) ‘Postscript to ‘Causation’ in Lewis, D. (1986b) Philosophical Papers Volume II. Oxford: Oxford University Press.
• (1986d) Counterfactuals. Cambridge MA: Harvard University Press. [Cited in Williamson 2007]
• (1999) ‘Elusive Knowledge’ in Papers in Metaphysics and Epistemology. Cambridge: Cambridge University Press.
• (2000). ‘Causation as influence’. Journal of Philosophy. 97. 182–97.
• (2004) ‘Void and Object’ in (eds.) Collins, J., Hall, N. and Paul, L.A. Causation and Counterfactuals. London: The MIT Press.

Lowe, E.J. (2002) A Survey of Metaphysics. Oxford: Oxford University Press.

Machamer, P, Darden, L and Craver, C.F. (2000) ‘Thinking about Mechanisms’ Philosophy of Science. 67. 1-25.

McCulloch, G. (1994) Using Sartre: An Analytical Introduction to Early Sartrean Themes. London: Routledge

Menzies, P. (2004) ‘Difference-Making in Context’ in (eds.) Collins, J., Hall, N. and Paul, L.A. Causation and Counterfactuals. London: The MIT Press.

• (2014) ‘Counterfactual Theories of Causation’ in (ed.) Zalta, E.N. The Stanford Encyclopedia of Philosophy. (Available online at: https://plato.stanford.edu/entries/causation-counterfactual/) [Accessed: 25/04/17]

Moore, A.W. The Evolution of Modern Metaphysics. Cambridge: Cambridge University Press.

Nolan, D. (2005) David Lewis. Chesham: Acumen Publishing.

Noonan, H. (2013) The Routledge Philosophy Guidebook to Kripke on Naming and Necessity. London: Routledge.

Schaffer, J. (2005) ‘Contrastive Causation’ The Philosophical Review. 114. 297-328.

• (2013) ‘Causal Contextualisms’ in (ed.) Blaauw, J. Contrastivism in Philosophy. London: Routledge.

Stalnaker, R. (2015) ‘Counterfactuals and Humean Reduction’ in (eds.) Loewer, B. and Schaffer, J. A Companion to David Lewis. Oxford: Wiley Blackwell.

Stoljar, D. (2015) ‘Lewis on Materialism and Experience’ in (eds.) Loewer, B. and Schaffer, J. A Companion to David Lewis. Oxford: Wiley Blackwell.

Williamson, T. (2007) The Philosophy of Philosophy. Oxford: Blackwell.

[1] Kripke did not mention Lewis by name when making this remark, but it is fairly easy to assume that was who he had in mind.

[2] Henceforth ‘Machamer et al’.