II. A FORMAL ARGUMENT

I will focus, for the most part, on a formal empirical SSA which has the structure normally attributed to actually expressed empirical SSAs. In presenting this formal argument, it will be useful to have in mind a simple model. Let us say that time is divided into two discrete periods, t ₀ and t ₁, where t ₀ is the current period and t ₁ lies in the future. Suppose that we – the members of some group (perhaps humanity as a whole) – must collectively decide in t ₀ whether to do Mild in t ₀. Then, having observed whether we do Mild in t ₀, our descendants will decide, in t ₁, whether to do Extreme in t ₁ (where Extreme is an extended or more extreme version of Mild). By ‘our descendants’, I mean simply those who will, in t ₁, be members of the group to which we, in t ₀, belong. With this basic model in mind, we can formulate our empirical SSA as follows:

• [EP] (Empirical Premise) By doing Mild in t ₀ we bring it about that our descendants will do Extreme in t ₁.Footnote ⁹

• [NP] (Normative Premise) We should not, in t ₀, bring it about that our descendants will do Extreme in t ₁.

• [NC] (Normative Conclusion) We should not do Mild in t ₀.

Many empirical SSAs that appear in the literature appear to lack an equivalent of NP, inferring an NC-like conclusion from an EP-like premise. However, an NP-like premise is clearly necessary if such arguments are to be valid. Hence, we can assume that where such a premise is not made explicit, it is nevertheless implicit.

III. A PROBLEM FOR THE FORMAL ARGUMENT

Assume, for the sake of argument, that EP is true. It follows, I will argue, that we have some evidence for the falsity of the other premise, NP.

According to EP, if we do Mild in t₀, then our descendants (our ‘post-Mild’ descendants) will do Extreme in t₁. But it might be thought that this fact – if it is a fact – counts as evidence that we should bring it about that our descendants do Extreme in t₁. After all, our descendants’ choice to do Extreme may well be based on a sincere belief that that is what they should do. And we often take the fact that other people sincerely believe some proposition as evidence for that proposition's being true. So perhaps the fact that our descendants would believe some proposition constitutes evidence for that proposition being true. We would thus have evidence that our descendants should do Extreme. Moreover, perhaps if they should do Extreme, we can infer that doing Extreme is objectively good and thus something that people should, other things being equal, promote. We would then have at least some reason to believe that we should bring it about that our descendants do Extreme: in doing so, we would be promoting an objective good and thereby acting as we should, other things being equal, act.Footnote ¹⁰

The thought I am entertaining here can be captured as follows:

I use ‘evidence’ in a weak sense whereby A is evidence for B just in case A provides a prima facie epistemic reason to believe B. This reason may be outweighed (it need not be a conclusive reason) and may also turn out to have no force whatsoever (it need not be a pro tanto reason).

The Evidence Thesis might, if true, throw some light on where the proponents of mistaken empirical SSAs have gone wrong. Consider the following argument advanced by a Protestant preacher in the United States prior to the abolition of slavery: we should not grant ‘colored men’ freedom from slavery, because this might cause us to grant them all sorts of other freedoms such that ‘a colored man might be the next governor; and colored men might constitute their Legislature, and sit on the bench as judges in their courts’.Footnote ¹¹ The implicit normative premise of this argument – that the further freedoms should not be brought about – is clearly false, and the proponent of the argument could presumably have accessed evidence for its falsity. According to the Evidence Thesis, one piece of evidence that he could have accessed was the fact (if it was a fact) reported by the empirical premise of his own argument: that abolishing slavery would lead future persons to grant African Americans those further freedoms. With the benefit of hindsight, it is clear that the granting of those further freedoms was the result of correct judgements.

The Evidence Thesis might also, if true, give us reason to doubt empirical SSAs that are not otherwise clearly mistaken. Thus, consider the empirical SSA in which ‘do Mild’ denotes ‘engage in gene therapy’ (that is, use genetic interventions to treat disease) and ‘do Extreme’ denotes ‘engage in genetic enhancement’ (that is, use genetic interventions to augment the capacities of disease-free humans). This is an argument against using gene therapy: EP claims that by engaging in gene therapy we will bring it about that our descendants engage in genetic enhancement, NP claims that we should not bring this result about, and NC concludes that we should not engage in gene therapy.Footnote ¹² The Evidence Thesis will then introduce the following doubt: if it is true that engaging in gene therapy would lead our descendants to genetically enhance themselves, then perhaps we have some reason to believe that they should engage in genetic enhancement, and, indirectly, that we should bring it about that they engage in genetic enhancement. And since it was not in any case obvious that we should not bring about genetic enhancement, this evidence may give us a decisive reason to drop the normative premise of the argument.

Of course, the Evidence Thesis will serve these diagnostic and critical purposes only if it is true, and I have hardly established that yet. We should consider the case for that thesis in more detail. That case, as I introduced it above, can be captured by the following three propositions:

Assuming that evidence is transitive (so that if A is evidence for B, which is evidence for C, then A is evidence for C), these three propositions jointly entail the Evidence Thesis. Thus, we can assess the truth of the Evidence Thesis by assessing the truth of each of these propositions. It is to this task that I now turn.

IV. PROPOSITION ONE

Proposition One is not obviously true. People often perform actions without also sincerely believing that they should perform those actions. We often act without even considering what we should do, and we sometimes act contrary to our normative beliefs (perhaps because we are weak willed, or the victims of coercion).

Nevertheless, acquiring the sincere belief that one should φ will generally increase a person's probability of ϕing. Thus, ϕing and sincerely believing that one should ϕ are to some extent positively correlated with one another. And it follows that, if our descendants would do Extreme, then we have a prima facie reason to attribute to them the belief that they should do Extreme. The fact that they would do Extreme constitutes evidence – though not necessarily decisive evidence – for their sincerely believing that they should do Extreme.

V. PROPOSITION TWO

Proposition Two maintains that the fact that our descendants would sincerely believe that they should do Extreme constitutes evidence that they indeed should do Extreme. There is some intuitive basis for this proposition. We often think that the beliefs of others constitute evidence for the truth of those beliefs. Adam Elga gives a nice example: you might well take the belief of a meteorologist that there is a 50 per cent chance of rain tomorrow as evidence for the truth of that belief.Footnote ¹³ Plausibly, we could extend the same line of thinking to future people with whom we would disagree.Footnote ¹⁴ But Proposition Two does not straightforwardly follow since there are also cases in which the sincere beliefs of others do not, intuitively, constitute evidence. For example, we would perhaps not take the sincere belief of a clairvoyant that the world will end in 2020 as providing any evidence for the proposition that the world will end in 2020. It would, however, be widely accepted that A should take B's belief that P as evidence for the truth of P where A is epistemically superior to B with regard to P. (I assume that a person's epistemic status is a function of her informational situation and her information-processing ability.) Arguably, Proposition Two will come out true if our post-Mild descendants would be our epistemic superiors on the question whether they should do Extreme. Whether Proposition Two would also come out true if our descendants were our epistemic equals is an interesting question, but not one that I can adequately address here.Footnote ¹⁵

In assessing the epistemic status of our post-Mild descendants, we have little information to go on. Some empirical SSAs may include premises according to which our post-Mild descendants will be our epistemic inferiors. It may be posited, for example, that their experience with Mild will corrupt them in some way (more on this in §VIII below). However, taken at face value, our formal argument – like most actually-expressed empirical SSAs – makes no such claim. The premises of that argument provide only the following information about our post-Mild descendants:

1. (a) They will, due to our doing Mild, do Extreme (from EP).

2. (b) With some non-zero probability, they will sincerely believe that they should do Extreme (from EP and Proposition One).

3. (c) They will live in a post-Mild world.

Fact (a) does not bear on the question whether our post-Mild descendants will be our epistemic superiors since it is consistent with their having either very high or very low epistemic status. We could try to explain (a) by supposing that our post-Mild descendants will be our epistemic inferiors. We could say, for example, that they will choose to do Extreme because their judgements will be clouded by an irrational attachment to Mild- and Extreme-like practices, born of their experience with Mild. However, we could equally explain (a) by supposing that our post-Mild descendants will be our epistemic superiors. Perhaps, for example, they will be in a better informational position than us to assess the risks associated with doing Extreme, and perhaps, from that superior position, they will judge that doing Extreme would be safe, and, all things considered, a good thing to do.

Fact (b) is also unhelpful. What we are faced with here is a case of intertemporal disagreement: we are supposing that we believe that our descendants should not do Extreme and that they believe that they should do Extreme. Is the fact of this disagreement evidence for the epistemic superiority of our descendants, or for our own epistemic superiority? Let us rule out, for simplicity, the unlikely scenario in which we and our descendants have precisely the same epistemic status. Thus, there are only two possibilities:

1. (1) We are epistemically superior to our descendants on the question whether they should do Extreme.

2. (2) Our descendants are epistemically superior to us on the question whether they should do Extreme.

It is clear that (1) and (2) are equally able to explain the observation that we and our descendants disagree. The fact of our disagreement does not favour one possibility over the other.

This brings us to fact (c): our post-Mild descendants will, by definition, live in a world in which Mild has already been done. Here, at last, we appear to have a fact that gives some clue as to the epistemic status of our descendants. Having experience with Mild, our post-Mild descendants will presumably be somewhat informed on the risks and benefits associated with doing Mild, and thus with extended or more extreme versions of Mild, such as Extreme. They will also, presumably, be aware of a greater range of arguments for and against both doing Mild and doing Extreme than we are, since our doing Mild is likely to stimulate debate about the ethics of these practices. Our descendants will thus almost certainly be in a superior informational situation to us regarding whether they should do Mild and this gives us prima facie reason to believe that they will be our epistemic superiors. This, in turn, gives us some reason to concur with their belief that they should do Mild, supposing that they indeed have such a belief.

VI. PROPOSITION THREE

If Propositions One and Two are true, then the fact that our post-Mild descendants would do Extreme constitutes evidence that they should do Extreme. I now turn to Proposition Three, which claims that the fact that our post-Mild descendants should do Extreme constitutes evidence that we should bring it about that they do Extreme. The thought underpinning this proposition was that, if our descendants should do Extreme, that must be because doing Extreme is objectively good, so that people should, other things being equal, promote it. It would follow that we should, other things being equal, bring it about that our descendants do Extreme. However, this assumes that if one should do some action, that action must be objectively good. But there are other ways of accounting for why one should do an action. Consider, for example, the following normative theory – a variant on what Derek Parfit calls the present aim theory:

Or alternatively, consider the following version of cultural relativism:

Both of these theories are what we might call ‘present relative’: they maintain that the contingent facts about what an agent should do at t are determined by facts about the world at t. Because they are present relative, they provide no support for Proposition Three. If we accept PA, then we cannot necessarily take facts about what our descendants should do as evidence for what we should do. According to PA, what our descendants should do in t₁ will be determined by their t₁ aims. But these t₁ aims will not determine what we should do. PA implies that only our t₀ aims can give rise to t₀ ‘shoulds’, and these aims may obviously be different from the t₁ aims of our descendants. Similar thoughts apply to CR. According to that theory, what our descendants should do will be determined by the norms that they endorse, and what we should do will be determined by the norms that we endorse. But since these norms could be different, there could be a gap between what they should do and what we should promote; it could be that they should do Extreme though we should not bring it about that they do this.

Present relative normative theories, such as PA and CR, are, however, controversial, and the premises of the formal argument give us no reason to accept them. (Indeed, the proponents of most empirical SSAs give no sign of accepting such theories.) At best, we could say that some present relative theory may be true. But an objective good theory, according to which the objective goodness of an act determines whether one should do it, might also be true.

Moreover, even if a present relative theory were true, the fact that our descendants should do Extreme would not constitute evidence against the view that we should bring it about that they do this. That our descendants should do Extreme would simply provide no evidence either way on what we should do, since the ‘shoulds’ are relative to different contingent facts.Footnote ¹⁷ Our situation, then, appears to be this. If a present relative theory were true, then the fact that our descendants should do Extreme would constitute no evidence either way on whether we should bring it about that they do Extreme. On the other hand, if an objective good theory were true, then the fact that our descendants should do Extreme would constitute evidence for the view that we should bring about their doing Extreme. But, in fact, we do not know which (or whether either) of these theories is true: we know only that either could be true and both have some plausibility. The question thus arises: given this uncertainty about which normative theory is true, which way does the evidence point? It seems clear that it supports the view that we should bring it about that our descendants do Extreme; we have identified one plausible theory according to which we have evidence for this view, and we have not identified any plausible alternative theory according to which we have contrary evidence. Proposition Three thus appears to be true.

VII. A FINAL OBJECTION TO THE EVIDENCE THESIS

Notwithstanding my arguments for Propositions One, Two and Three, it might be thought that one of these must be false since the Evidence Thesis is itself clearly false. One reason for thinking this is that the Evidence Thesis appears to focus illicitly on the t₁ world that will come about if, in t₀, we do Mild. The Thesis holds that what our descendants should do in the post-Mild t₁ world constitutes evidence for what we should do in t₀. But surely there is another possible t₁ world that we need to consider: the t₁ world that will come about if we do not do Mild in t₀. In that world (the ‘no-Mild’ t₁ world), our descendants might not do Extreme. To be consistent, then, we must surely take this fact as evidence that we should not bring it about that our descendants do Extreme. And this evidence would precisely offset the contrary evidence cited by Evidence Thesis.

Two things can be said in response to this objection. First, we need not assume that our no-Mild descendants would, with a probability of one, refrain from doing Extreme. The formal argument does not claim this. It is perhaps implicit in the argument that our post-Mild descendants will be more likely to do Extreme than our no-Mild descendants. But it remains possible, for example, that our no-Mild descendants would be as likely to do Extreme as not. And whether they do Extreme would then provide no evidence either way on whether they should do Extreme, nor on whether we should bring it about that they do so. We would thus have no evidence to offset the evidence cited by the Evidence Thesis.

Second, even if our no-Mild descendants would choose not to do Extreme, this might not provide any evidence for what they or we should do. This is because our no-Mild descendants might not be our epistemic superiors on the question of whether they should do Extreme. I said above that our post-Mild descendants would, plausibly, be our epistemic superiors on this question, but that assessment was based on the assumption that they would have some experience with Mild. Our no-Mild descendants would have no such experience, and we therefore have little or no reason to believe that their informational situation would be any better than our own. The analogue of Proposition Two may thus not hold for our no-Mild descendants.

VIII. IMPLICATIONS

I have argued that,

That is, the fact reported by EP provides a prima facie reason to believe that NP is false. We should now consider the implications of this thesis for some actual empirical SSAs.

It should be noted at the outset that some empirical SSAs are immune to the Evidence Thesis. These are the arguments whose normative premises are so well supported that our justification for accepting them could not be undermined by the sort of counter-evidence cited by the Evidence Thesis. An example of such an argument might be the empirical SSA in which ‘do Mild’ denotes ‘engage in voluntary euthanasia’ and ‘do Extreme’ denotes ‘engage in involuntary euthanasia’. The normative premise of this argument is that we should not bring it about that our descendents engage in involuntary euthanasia. Arguably, we have such strong reasons to believe this premise that the countervailing reason captured by the Evidence Thesis could not be decisive.

Indeed, it is doubtful whether the Evidence Thesis would have any dialectic impact at all in discussion of the euthanasia SSA. I imagine that the Evidence Thesis might be cited by someone who accepted the empirical premise of an empirical SSA but nevertheless wanted to resist that argument's conclusion. But arguably, no-one would accept the empirical premise of the euthanasia SSA without already taking the Evidence Thesis to be false. This is because the normative premise of that argument is so plausible that both proponents and opponents of the argument are likely to take it as common ground: it will simply be granted that we should not bring about involuntary euthanasia. Once this is granted, it is difficult to see how one could come to accept the empirical premise of the argument – that our engaging in voluntary euthanasia will bring it about that our descendants engage in involuntary euthanasia – without already accepting that our engaging in voluntary euthanasia will somehow corrupt our descendants (either rendering them our epistemic inferiors, or preventing them from regulating their behaviour on the basis of sincere moral judgements). How could we bring it about that our descendants engage in voluntary euthanasia except by corrupting them in one of these ways? It seems, then, that to accept the empirical premise of this SSA, one must accept that our descendants will be corrupted. But this undermines the Evidence Thesis, since if our descendants will be corrupted, then we probably have no reason to take the fact that they would engage in voluntary euthanasia as evidence for what we should do.

Note, however, that when empirical SSAs contain obviously true normative premises – as in the euthanasia SSA – they are often particularly susceptible to traditional critiques which target their empirical premises. If it is obvious that we should not bring about some practice, this will often be because that practice is obviously objectively immoral. But if that is so, we have grounds to hope that our descendants will recognize its objective immorality and abstain from it. The empirical premises of empirical SSAs are frequently most plausible when the feared practice is plausibly but not obviously immoral, and when the normative premise of the argument is thus not obviously true. But it is in precisely these arguments that the Evidence Thesis comes into its own. Consider, for example, the argument against gene therapy mentioned earlier. In this argument, ‘do Mild’ denotes ‘engage in gene therapy’ and do Extreme denotes ‘engage in genetic enhancement’. The normative premise asserts that we should not bring it about that our descendants engage in genetic enhancement. But this is a controversial claim. It is not obviously true, and it therefore seems quite likely that evidence of the sort cited by the Evidence Thesis could give us decisive reason to drop that premise.

Moreover, in the gene therapy SSA, the Evidence Thesis is not dialectically impotent as it was in the euthanasia SSA. Since the normative premise of the gene therapy argument is uncertain, it may not be granted from the outset by all discussants. (Indeed, proponents and opponents of the argument may be more likely to agree on the empirical premise.) It therefore remains open to opponents of the argument to accept the empirical premise of the argument without rejecting the Evidence Thesis: one could accept that our engaging in gene therapy will cause our descendants to engage in genetic enhancement without believing that it would corrupt them. The opponent could then adduce the empirical premise of the argument and the Evidence Thesis against the normative premise.

It thus seems that, when directed against empirical SSAs with uncertain normative premises, appeals to the evidence thesis might be both decisive and dialectically potent. Aside from the gene therapy SSA, other empirical SSAs in this category might include the following:

• • We should not introduce gun registration, since this would lead our descendants to engage in gun confiscation.

• • We should not permit pharmaco-genetic testing, since this would lead our descendants to engage in widespread genetic testing for many conditions.

• • The state should not provide parents with school vouchers that can be redeemed at private secular schools since this would lead its descendant-state to provide vouchers that can be redeemed at private religious schools.Footnote ¹⁸

• • We should not engage in research using human embryonic stem cells, since this would lead our descendants to engage in reproductive cloning.Footnote ¹⁹

• • We should not permit the police to install video cameras on street lamps since this would lead our descendants to permit the permanent archiving of tapes from these cameras.Footnote ²⁰

As far as I can see, proponents of such arguments can avoid my critique altogether only by inserting into their arguments a further premise according to which our descendants will be corrupted by our action.Footnote ²¹ In some cases, there may be good reasons to accept this premise. Consider, for example, the following empirical SSA: I should not smoke one packet of cigarettes, since doing so would lead to my becoming a regular smoker and I should not bring it about that I am regular smoker. Assuming that the empirical premise contained in this argument is true, it seems that I have some reason to believe that my future self will be my epistemic inferior, or that he will be incapable of guiding his smoking behaviour on the basis of sincere normative beliefs. This is because, as it happens, the best explanation for the truth of the empirical premise of this argument in this case is that tobacco is addictive. And addiction is, arguably, a paradigmatic example of a corrupting process (that is, a process which reduces epistemic status or overrides sincere normative beliefs).Footnote ²²

However, there are many empirical SSAs for which this additional premise may be less plausible. Consider again the gene therapy SSA. Arguably, the best explanation for the truth of the empirical premise of this argument would simply be that, as a matter of empirical fact, technologies used for therapeutic ends tend, after a time, to be adapted for use as enhancements. But this explanation has no bearing on the question whether our descendants will be corrupted or not, since it makes no assumption about why therapy tends to develop into enhancement. Consider alternatively the argument that we should not permit the police to install video cameras on street lamps since this would lead our descendants to permit the permanent archiving of tapes from these cameras. Perhaps the best explanation for the truth of the empirical premise of this argument would be that, having experience with the use of video cameras, our descendants will come to recognize the advantages of archiving the tapes from those cameras. This explanation would undermine, rather than support, the new premise that our descendants will be corrupted.

We appear to have identified, then, a group of empirical SSAs that may be susceptible to criticisms based on the Evidence Thesis: namely, those which have uncertain normative premises. To the extent that we have reason to accept the empirical premises of those arguments, we have potentially decisive prima facie reason to reject their normative premises. And to avoid this difficulty, we need to add further premises – asserting that our descendants will be corrupted – which are sometimes implausible.

Thus, in order to assess comprehensively any empirical SSA, we should, in addition to assessing its empirical premise, consider (1) whether the argument is susceptible to the Evidence Thesis by determining whether the normative premise is uncertain. If it is susceptible, we should consider (2) whether the argument can be rescued from the Evidence Thesis by inserting a plausible premise to the effect that our doing Mild is likely to corrupt our descendants.Footnote ²³