Philosophical Pluralism and Modest Dogmatism

Philosophers are sometimes prone to excessive skepticism, especially in the face of persisting disagreement. People often seem really bothered by the lack of consensus in philosophy (including, e.g., Derek Parfit, Jason Brennan, and most recently, Liam Kofi Bright).  But a large portion of such worries seem to stem from a failure to appreciate when actual disagreement is (distinctively) undermining:

In cases of what we might call 'non-ideal' disagreement, there's a presumption that the disagreement is rationally resolvable through the identification of some fallacy or procedural mis-step in the reasoning of either ourselves or our interlocutor. The disagreement is 'non-ideal' in the sense that we're only disagreeing because one of us made a blunder somewhere. We are sufficiently similar in our fundamental epistemic standards and methods that we can generally treat the other's output as a sign of what we (when not malfunctioning) would output. The epistemic significance of the disagreement is thus that the conflicting judgment of a previously-reliable source is some evidence that we have made a blunder by our own lights, though we may not yet have seen it.

Many philosophical disagreements do not have this crucial feature.  This is because "(i) there are many possible internally coherent worldviews, (ii) philosophical argumentation proceeds through a mixture of ironing out incoherence and making us aware of possibilities we had previously neglected." As a result, many philosophical disagreements simply reflect different substantive starting points rather than any purely procedural blunder.  And the fact that somebody exists who holds different substantive starting points than you has zero epistemic import over and above the prior observation that there are coherent alternatives to your own view (which you really should already know!).

Now, it's totally fair to worry about the epistemic significance of coherent alternatives. As I put it previously, it follows that "(iii) even the greatest [philosophical] expertise... will only help you to reach the truth if you start off in roughly the right place. Increasing the coherence of someone who is totally wrong (i.e. closer to one of the many internally coherent worldviews that is objectively incorrect) won't necessarily bring them any closer to the truth."

Good reasoning provides no guarantee of truth. There's a real possibility that we're irreparably mistaken, in which case no amount of procedurally conscientious reasoning would see us right.  Moreover, there's no neutrally-recognizable standard by which we can determine whether we are irreparably mistaken in this way. (If there were, the error wouldn't be so irreparable, after all!)  These facts may be disheartening to those who hoped for a transparent light of Reason to guide our way; but epistemic maturity requires us to recognize that such Cartesian hopes were never really reasonable or realistic in the first place.  We muddle by as best we can, and hope for the best.  If we successfully reach (or at least approximate) an internally-coherent position, it's possible that the one we reached is the one true view of the matter, but we cannot expect to be able to prove this to any who doubt us -- not even ourselves.  The most we can hope for is to be, in a sense, philosophically lucky. (But that's a fine thing to hope for!  Nothing is gained by holding out for unattainable Cartesian certainties.  It remains well worth striving for coherence, since that at least gives us a chance at being right.)

Once the above lessons are properly internalized, viewpoint diversity and philosophical dissensus comes to seem entirely appropriate. We need advocates to work out all the options, after all. (Nothing is gained by having a coherent alternative view be routinely ignored or overlooked: a merely sociological uniformity of opinion is no "consensus" worth having.) And if a view is internally coherent, you shouldn't expect to be able to argue a sophisticated advocate out of that position. Nothing forces them to share your premises!

Rather than seeing this as some deep failure of "analytic philosophy" (as if different training would somehow break the logical symmetry between modus ponens and modus tollens), I'd encourage clear-eyed acceptance of this reality, combined with default trust or optimism about your own philosophical projects. After all, unlike all those fools who disagree with you, YOU'RE beginning from roughly the right starting points, right? ;-)

Lessons from the Pandemic

It's generally recognized that our (American) response to the Covid-19 pandemic was disastrous. But I think far fewer appreciate the full scale of the disaster, or the most significant causal levers by which the worst effects could have been avoided.  (Yes, Trump was bad.  But his public health disinformation and politicization of masking -- while obviously bad -- may prove relatively trivial compared to the mammoth failings of our public health institutions and medical establishment.)  Much of the pandemic's harm could have been mitigated had our institutions been properly guided by the most basic norms of cost-benefit analysis.  Consider:

What Makes Your Papers Worth Reading?

Given how many academic papers are out there, it would be useful to have more filtering and discovery mechanisms for helping us to find the ones we might be most interested in.  One thing that could help is if authors themselves offered a concise 'overview' of what they think makes their various papers worth reading (when they are).  Many of us already list our papers on our websites, but (i) standard academic abstracts rarely do a good job of explaining why a paper is worth reading, and (ii) who reads academic websites anyway?  So I'm going to take a stab at doing this in a blog post, and invite others to follow suit (whether on Facebook or wherever you like: feel free to additionally post your response in the comments here, especially if your research interests overlap with mine at all). What lessons from your work do you wish were more widely appreciated?

Ordered by how much I happen to like each paper today:

Singer's Pond and Quality of Will

Singer argues that, just as we're obliged to save a drowning child at modest cost to ourselves (e.g. ruining an expensive suit), so we're obliged to help the distant needy when we're in a position to do so (e.g. by donating to GiveWell-recommended aid organizations). People often balk at this comparison, but I don't see any plausible grounds for escaping the conclusion that we have similarly strong reasons to act in either case.

What's more, I don't think this particular result is really all that counter-intuitive, either. Of course we have incredibly strong reasons to save innocent lives whenever we can! What could be more important, or more worth choosing, than that? This claim about the strengths of various reasons for action -- call it the Act Evaluation -- is eminently plausible.

What is counter-intuitive, I think, is the putative implication that when we fail to donate to effective charities we are thereby just as bad, or as blameworthy, as a person who lets a child drown before their eyes.

The Separateness of Persons: Commensurability without Fungibility

It seems to me that the famed "separateness of persons" objection to consequentialism rests on the confused assumption that commensurable values (ones that can be compared and traded off against each other) are thereby fungible values (such that a loss to one is not merely outweighed, but actually cancelled, by a greater gain to another). I'll explain in a moment why this is a mistake. But first, let's motivate the objection with a simple case:

Connie has just enough anti-venom to save one of the two poison victims before her. Now, faced with their pleading faces, but realizing it makes no difference to the total welfare, Connie finds herself totally uninterested in the question of who to save. It strikes her as no more normatively significant than the choice between a $20 bill or two tens.

Satisficing by Effort

Satisficing Consequentialism aims to capture the intuitive idea that we're not morally obligated to do the best possible, we merely need to do "good enough" (though of course it remains better to do better!). Ben Bradley, in 'Against Satisficing Consequentialism', argues convincingly against forms of the view which introduce the baseline as some utility level n that we need to meet. Such views absurdly condone the act of gratuitously preventing boosts to utility over the baseline n. But I think there is a better form that satisficing consequentialism can take. Rather than employing a baseline utility level, a better way to "satisfice" is to introduce a level of maximum demanded effort below which one straightforwardly maximizes utility. That is:

(Effort-based Satisficing Consequentialism) An act is permissible iff it produces no less utility than any alternative action the agent could perform with up to X effort.

The Puzzle of the Self-Torturer

Warren Quinn's (1993) 'Puzzle of the Self-Torturer' can be described as follows:

Suppose someone — who, for reasons that will become apparent, Quinn calls the self-torturer — has a special electric device attached to him. The device has 1001 settings: 0, 1, 2, 3, …, 1000 and works as follows: moving up a setting raises, by a tiny increment, the amount of electric current applied to the self-torturer's body. The increments in current are so small that the self-torturer cannot tell the difference between adjacent settings. He can, however, tell the difference between settings that are far apart. And, in fact, there are settings at which the self-torturer would experience excruciating pain. Once a week, the self-torturer can compare all the different settings. He must then go back to the setting he was at and decide if he wants to move up a setting. If he does so, he gets $10,000, but he can never permanently return to a lower setting. Like most of us, the self-torturer would like to increase his fortune but also cares about feeling well. Since the self-torturer cannot feel any difference in comfort between adjacent settings but gets $10,000 at each advance, he prefers, for any two consecutive settings s and s+1, stopping at s+1 to stopping at s. But, since he does not want to live in excruciating pain, even for a great fortune, he also prefers stopping at a low setting, such as 0, over stopping at a high setting, such as 1000.

This is generally framed as a puzzle for rational choice: what setting should the self-torturer rationally select? But I think it is illuminating to first consider the question of objective value: which outcome would in fact be best for the agent?

Personal Concern and Chains of Counterparts

In chapter 9 of his forthcoming book The Limits of Kindness (cited with permission), Caspar Hare offers an ingenious argument for the view that transitivity + minimal decency (i.e. preferring pareto-superior alternatives, whereby some people are better off and nobody is worse off) actually commits us to anonymous benevolence, i.e. preferring world A over world B whenever the inhabitants can be put into a 1:1 correspondence such that the person in A is sometimes better off, and never worse off, than their paired member of B. So, for example, it commits us to preferring (contra Taurek) that we save the many rather than the few in rescue cases, and that we bring into existence better-off rather than worse-off people in non-identity cases. Very roughly, it's an argument that moves us from personal to impartial concerns.

Consequentialist Agents: Fittingness and Fortune

[This is something of a manifesto for my current research project...]

Critics of consequentialism often object to how a consequentialist agent would (allegedly) think. They claim that the consequentialist agent is, in some sense, a bad character. Defenders of consequentialism typically dismiss such objections by citing the distinction between 'criteria of rightness' and 'decision procedures'. (Utility provides the criterion that determines the moral status of an act; it's a further question whether agents ought to attempt to calculate utilities themselves.) This is not entirely satisfactory. There remains a real objection here that needs to be addressed, not just dismissed. As I will explain, consequentialists still need to say something about what a 'rational' or fitting moral (consequentialist) agent would look like -- and when they do, this leaves room for others to object that the agent thus pictured is not in any sense morally 'rational' or non-instrumentally ideal.

Reasons Deflate Global Consequentialism

I'm beginning to think that global consequentialism isn't really a distinctive position in its own right. It may be better understood simply as act consequentialism that takes care to not make straightforwardly false claims about the consequential value of other evaluands (like rules, motives, etc.). Global consequentialists recognize that these other evaluands might bring about good or bad consequences by "external means" other than the agent's own downstream actions. But the denial of this is not anything so principled as an opposing position. It's just an oversight.

Here's the structure of my argument. I take as a premise the assumption that substantive normative claims must fundamentally concern normative reasons (of one kind or another). The core of my argument thus hinges on identifying what categories of practical reasons there are, about which consequentialist theories may then make substantive claims. When we do this, I argue, we find that there is nothing more for global consequentialism to be besides act consequentialism that avoids silly oversights.

The Mark of the Instrumental

How do we distinguish intrinsic (final) value from merely instrumental value? Simon Keller argues that relationships are of merely instrumental value on the grounds that (i) unhappy relationships which lack all instrumental value seem to have no value at all; (ii) we don't seem to have any reason to create new relationships (e.g. between happy hermits) that would have no further beneficial consequences; and (iii) the badness of bad relationships is typically explained reductively ("it's making you both miserable"). Even granting these claims for sake of argument, I'm not convinced it follows that relationships are of merely instrumental value. That would imply that they are fungible; but a value could be contingent in all the ways specified above without thereby being fungible.

For clarity, let's translate value-talk into reasons-talk (cf. "buck passing"), such that 'X is valuable' becomes 'X is desirable', i.e. there are reasons to want X. If X has merely instrumental value, then the reasons to want X are wholly parasitic on our reasons to want some other Z (to which X is conducive). That is, we have no reason to want X over and above our reasons to want Z. Merely instrumental value thus consists in there being reasons only for instrumental desire. Reasons to take something as an ultimate end, by contrast, amount to "intrinsic" or final value.

Now, why think that instrumental value must be fungible? Intuitively: one means is as good as any other, so long as it comes to the same in terms of our final goals. If we have no reason to want X over and above our reasons to want Z, then the fact that the latter goal is served just as well by some alternative means Y (without detrimentally impacting our other goals) entails that we can have no reason to prefer X to Y. It's just like swapping two $10 bills for a $20.

But it seems bizarre to think that relationships are fungible in this way. Most people would not wish to swap spouses with a stranger, even if they knew this would have no effect on how happy (etc.) each person was. Even more dramatically, suppose that one's spouse was long ago kidnapped and secretly replaced by a cleverly disguised robot. (The real person is then set up in a duplicate world with a cleverly disguised robotic copy of you, etc.) This is clearly a terrible outcome. But all the instrumental benefits remain as before: each robospouse provides their human partner with happiness, and promotes their moral flourishing, etc. What's missing is the genuine personal connection. Since the loss of it makes no instrumental difference, the value here must be intrinsic.

What, then, should we make of Keller's three arguments? I think they rest on a failure to distinguish the content of a reason and its preconditions. This is clearest in the case of (i) and (iii). Contingency of desire does not imply instrumentality -- as explained in my old Railton-inspired post on 'Regulating Aims' -- and likewise for desirability. Some X may be non-instrumentally desirable only if it leads to Z. Given the absence of fungibility here, this is not the same as X being desirable merely as a means to Z.

(We may also distinguish having Z-contingent or 'conditional' reasons to desire X straight out, versus having straight-out reason to conditionally desire X given Z. But since neither of these is merely instrumental, the distinction is not essential for present purposes.)

So the mere fact that instrumentally worthless ("unhappy" for short) relationships have no value at all does not imply that the value of other relationships is merely instrumental. It may be that only if a relationship is conducive to happiness do we have reason to want to maintain it; but nonetheless, what this reason mandates is a non-instrumental desire to maintain this very relationship, and not just any old means to equal happiness.

Of course, if the reason were contingent in this way, that would also suffice to make sense of why one could reductively explain the undesirability of the relationship by citing the absence of this precondition. "This relationship is bad because it is making you both miserable" implies that if it were happier then you'd have reason to desire the relationship. It does not imply instrumentality (or that you would have reason to desire the relationship only for the happiness etc. it would bring): again, a contingent reason might just as well mandate a non-instrumental desire.

So (i) and (iii) are invalid grounds for inferring instrumentality. This post is too long already, so I'll leave discussion of (ii) for the next post...

On Rosen's MR-Skepticism

[I might end up expanding the following sketch into a real paper at some point, so any feedback would be most welcome...]

In 'Skepticism About Moral Responsibility', Gideon Rosen argues for the view that confident positive judgments of blameworthiness are (almost) never warranted. The argument proceeds in two parts. First he argues that blameworthiness can only originate in akratic action, whereby the agent acts in full knowledge of their wrongdoing (and its normative significance). In the second part, Rosen suggests that we rarely have sufficient grounds for confidently attributing such full-blown akrasia to agents who have acted wrongly. I will focus only on the first part of the argument, since this is where the main philosophical action takes place.

Why think that blameworthiness can only originate in akratic action? Rosen's argument is quite intricate, but the core idea is that non-akratic wrongdoing is merely done “from ignorance”, and is thus excused unless the agent is to blame for her state of ignorance. But if the agent also acted from ignorance in bringing about this prior state, then that act will likewise be excused, unless ... etc. The agent's excuses will only run out if we can stop the regress at a point where she acted in full knowledge of her normative error: that is, if she acted akratically.

More precisely: Rosen begins by distinguishing derivative and original responsibility for an act or event. For example, if Jekyll intentionally takes a pill that makes him berserk, then he is (derivatively) responsible for the resulting damage in virtue of being (originally) responsible for the negligent act of taking the pill. Rosen then claims that “an action done from ignorance is never a locus of original responsibility.” Let's call this the RAID principle (responsibility for actions done from ignorance is merely derivative). Rosen supports this principle by appeal to ordinary examples of factual ignorance, such as accidental poisoning. If someone secretly plants arsenic in my sugar bowl, it is not my fault that I end up poisoning you. But if I routinely store arsenic in unlabeled sugar bowls in my pantry, this prior negligence puts me back on the hook for causing the accident. Crucially, Rosen takes the RAID principle to be fully general, applying to normative as well as factual ignorance.

The RAID principle tells us that an act done from ignorance cannot be a locus of original responsibility. But any agent who non-akratically acts wrongly thereby acts from ignorance of some kind: either they aren't aware of the factual considerations that make the act wrong, or they aren't aware of the normative significance of those facts: that they make the act wrong, and that they thereby count decisively against so acting. If an agent were to know all these facts and yet still perform the act, then their act would qualify as 'akratic' by Rosen's definition. RAID thus straightforwardly implies that blameworthiness can only originate in akratic action. That is: for X to be blameworthy for A, A must either itself be an akratic act (if X is originally responsible for A), or else be the upshot of an akratic act (if X is derivatively responsible for A). To avoid this conclusion, we will need to undermine RAID.

In order to see why RAID is false, we need to clarify our understanding of 'original' and 'derivative' responsibility, which will bring to light two very different kinds of culpability in ignorance. In particular, I will distinguish 'diachronic' and 'synchronic' cases of culpable ignorance, and argue that RAID holds only of the former. The intuitive plausibility of the principle derives from neglecting the latter class of cases, which – we will see – provide ample intuitive counterexamples to the RAID principle. I will close by considering what Rosen might say about these neglected cases.

Rosen's loose talk of 'original' and 'derivative' responsibility suggests a picture on which responsibility is a two-place relation, between an individual and the acts or outcomes that they are “responsible for”. But this occludes some important structure which I want to bring to the fore. Responsibility is better understood as a three-place relation between an agent, an act of theirs (which I will call the 'focal act'), and an event (or set of events) for which the agent's action renders him liable. Intuitively: we are responsible, in acting a certain way, for various outcomes. The focal act is the exercise of agency in virtue of which we become liable for praise or blame. That is to say: it is the locus of original responsibility. But we can now see that there is no such thing as a locus of 'derivative responsibility', so that old distinction seems inapt (or at least potentially misleading). Rosen would have us say that Jekyll the pill-popping berserker is derivatively responsible for vandalism due to being originally responsible for pill-popping, but the “responsible for” locution is used here equivocally. It would be clearer just to say that Jekyll is responsible, in taking the pill, for the subsequent damage.

This makes it clear that (in the described case) there is only one occasion of Jekyll exercising his agency in a blameworthy manner. His berserk vandalism is not itself a blameworthy act (assuming it is an act at all), it is merely an event for which his blameworthy act of pill-popping renders him liable. (Notice that he would still be blameworthy in exactly the same way even if he had slipped the pill to an innocent third party who then went berserk in the china shop. The innocent person's berserk vandalism isn't blameworthy, and neither is Jekyll's. It's his pill-popping, in either case, that is the sole blameworthy action, though it makes him liable for downstream events – including the event of himself or others acting in berserk fashion.)

Rosen may grant all this, but ask why the terminological precision matters. Well, for one thing, it helps us to see more clearly what sorts of considerations are eligible to excuse the agent. Sometimes ignorance is a legitimate excuse, but only if it is concurrent with the focal act. It would clearly be absurd for an agent to plead innocence on the grounds that, despite knowing full well the risks of their action as they performed it, they later forgot this (by the time the risked outcome actually eventuated). And this remains true even if the risked outcome in question was another act of theirs. Granted, their ignorance – like Jekyll's incapacity – might excuse the downstream act from qualifying as a second occasion of blameworthiness. But it obviously does nothing at all to excuse the earlier “benighting” act (to borrow Holly Smith's term), or to cast doubt on whether the “unwitting” outcome falls within its scope of liability (that is, as an event for which the agent, in performing the benighting act, became liable). So, in many of Rosen's examples, the kind of 'ignorance' involved – whether culpable or not – was never even prima facie eligible to serve as an excuse, at least insofar as we are treating the earlier act as focal. To properly understand the excusing potential of ignorance (and its limits), we need to instead focus our attention on cases of ignorance concurrent with the focal act or exercise of agency that is to be assessed.

Let's return to the original case of the accidental poisoning. We can imagine two very different ways in which an agent might be 'culpable'¹ for their failure to recognize the risk of arsenic poisoning. In one case, the culpability is wholly 'diachronic': suppose I planted the arsenic myself, and then took a pill that wiped my memory of this event. My subsequent decision to spoon white powder (now reasonably believed to be sugar) into your tea is itself an intrinsically innocent, blameless act. As in the Jekyll case, it is only my earlier 'benighting' action that is blameworthy (and renders me liable for the eventuating damages). This sort of case supports the RAID principle, which effectively claims that actions done from ignorance are never themselves blameworthy [qua act], but are at most events which some earlier blameworthy act may render us liable for.

Alternatively, we can imagine a case of 'synchronically' culpable ignorance, where my unwitting poisoning no longer seems even intrinsically wholly innocent. Suppose my roommate left the arsenic in a labeled bowl in the pantry. When I grab what looks like a sugar bowl, my eyes flit briefly over the label ('arsenic'), without fully taking it in or consciously registering the significance of this. Further suppose that the reason why I don't fully process the visual evidence available to me is just because I don't really care very much about you. Maybe I even want you dead, on some subconscious level. In any case, we can stipulate that there is evidence available to me that the bowl contains arsenic, and my ignorance depends upon my callous disregard for your welfare. If I cared more, as I morally ought to – or if I had all the beliefs that I rationally ought to, given my evidence – then I would never have so thoughtlessly spooned the white powder into your tea. I didn't appreciate the risk at the time, but I should have,² and so in this sense my ignorance is 'synchronically culpable'. Intuitively, in this sort of case, I am still blameworthy in acting so thoughtlessly. Unreasonable ignorance does not (fully) excuse.

We thus have a counterexample to the RAID principle. It may be true that acts done from merely diachronically culpable ignorance, being intrinsically (synchronically) reasonable, are never blameworthy. But, intuitively, it looks like we can be blameworthy – 'originally responsible' – in acting from 'synchronically culpable' or unreasonable ignorance. So the fully general RAID principle lacks intuitive support after all, leaving us with no reason to think that blameworthiness must originate in akratic action. It can just as well originate in unreasonable action. And note that even if we're rarely in a position to attribute akrasia to wrongdoers, we're very often in a position to know that they acted unreasonably. Hence Rosen's argument fails.

At this point, Rosen might respond by giving up his claim to ordinary case-based intuitions, and instead seeking to defend RAID on purely theoretical grounds. As he observes in a footnote, it's one thing to say that a person “should have known better”, and another to say that they are culpable for this failure. So he at least has the theoretical resources to coherently reject the commonsense view gestured at in my previous paragraph. It just remains to motivate his more radical view. As with most skeptical arguments, this will require appealing to intuitive-sounding abstract principles (rather than cases). Here are two that together imply RAID:

1. “If X does A from ignorance, then X is culpable for the act only if he is culpable for the ignorance from which he acts.”
2. “X is culpable for failing to know that P only if his ignorance is the upshot of some prior culpable act or omission.”

Why believe (2)? It follows from the general principle that only acts (including mental acts, but excluding internal 'occurrences' with respect to which we are 'passive') are eligible to serve as loci of responsibility. Since ignorance isn't an exercise of agency – something you do – it can't be a site of blameworthiness. At most, it can be an outcome that some prior act made you liable for. (It may also indicate that one is vicious, much as a brain scan might, but to be a bad person is not yet to be blameworthy, since the latter kind of evaluation concerns what you've done.)

Suppose we grant all that. Why would we then accept (1)? It might be confused with the more plausible claim that reasonable ignorance excuses, and hence that an act done from ignorance is blameworthy only if the agent was unreasonable in believing as they did. But (1) is a much stronger claim since, we are supposing, even unreasonable ignorance can't originate blameworthiness. Even so, we may still think that an act done from unreasonable ignorance, manifesting a lack of good will (in case of motivated ignorance), may yet serve as a locus of blameworthiness. If we think this is possible – as certainly appears to be the case in the 'synchronic' cases I've highlighted – then we will reject (1).

Our final assessment of skeptical arguments typically depends on whether we are more inclined to trust intuitions about cases or abstract theoretical principles (e.g. that knowledge requires certainty, that an epistemic [e.g. inductive] rule cannot justify itself, etc.). In this instance, though premise (1) may be prima facie plausible-sounding, I think that once we see understand what it actually claims, and how radically opposed it is to our ordinary case-based intuitions, we should simply reject it as unmotivated.

1In a broad sense; we'll later see that this isn't quite the right term.

2Limitations of space prevent me from further exploring whether it is the moral or the epistemic vice that is relevant here. I will use the term “reasonable” broadly to cover both kinds of normativity. I should also mention that the alternative label 'synchronic culpability' turns out, as explained below, to be something of a misnomer.

Understanding (Zombie) Conceivability Arguments: Part II

So far we have established that a legitimate conceivability-possibility inference must start from the ideal conceivability of a semantically neutral statement. The final requirement I want to discuss is that the statement to be conceived should be specified in transparent, uncontested terms.

What do I mean by a 'transparent' specification? I mean to exclude broad general terms (like 'physical' or 'non-physical') whose extensions might be controversial. This helps us to avoid pointless terminological debates. Note that it's a bit sloppy to argue, "Conceivably a world could be physically just like ours but lack consciousness, therefore physicalism is false," because it is unclear exactly what we are meant to be conceiving: the phrase 'physically just like ours' is too opaque: different people might take it to mean different things. [Plus, its invocation of actuality - 'our world' - technically violates the semantic neutrality requirement.] So, for maximum rigor, it should be replaced by a precise microphysical description P that - without making any explicit reference to qualia - is accepted by the physicalist as true of our world. We can then argue from the conceivability of (P & ~Q) to the precise conclusion that the qualitative properties in Q are not reducible to the properties specified in P, and so must be explicitly added as further primitives to the base facts.

It is important to note the precise conclusion here. It is not that Q is non-physical, or that physicalism is false. It is merely that Q does not reduce to anything included in P. Importantly, none of the premises say anything about whether P exhausts the physical properties. (We chose P so that the physicalist should agree that it does exhaust the physical properties, but that occurs outside of the scope of the argument itself. This is a vital point that I will return to.) So one could, in principle, respond that Q is a primitive "physical" property that doesn't reduce to any of the other ones listed. This "physicalist" will insist that a full specification of the physical base facts P* must include Q as a basic conjunct, in which case (P* & ~Q) will be straightforwardly inconsistent, seeing as how it would be equivalent to ((P & Q) & ~Q) for some P. Clearly, this "physicalist" is completely immune to refutation by the kind of conceivability argument discussed here.

That's fine, except that this is to be a "physicalist" in name only. Once you grant that qualia are primitive, and must be explicitly included in the base facts, that's all the dualist cares to establish. Whether you call these primitive mental facts "physical" or "non-physical" is mere semantics.

The target of the zombie argument is instead the strict physicalist, who refuses to countenance primitive mental properties. He believes that we can fully specify the base facts of the world without any explicit mention of qualia. P alone will suffice for Q, on his view, just as P suffices to fix the macroscopic table and chair facts. But since P (unlike P*) makes no explicit mention of Q, we have some chance of running a conceivability argument against this view, since it looks like (P & ~Q) could be conceivable. It's not a knock-down argument, of course: one could reasonably deny that it really is ideally conceivable at the end of the day. But at least the zombie argument has a chance, unlike in the previous case.

Why am I harping on about this at such length? Well, I think this background can help us diagnose the misunderstanding implicit in Richard Brown's dismissive parody:

I am conceiving of a world that is just like this one in all non-physical respects except that it lacks consciousness. Therefore dualism is false.

This exhibits the same sloppiness exhibited by the sloppy version of the zombie argument discussed at the very start of this post. The difference is that while the zombie argument can (as demonstrated) be made more precise, this parody -- what RB now calls his "zoombie argument" -- cannot.

What's the problem? Well, as before, it's entirely opaque what we are supposed to be conceiving, since there is no uncontested specification of the "non-physical respects" of the world that we can plug in to the argument. I take it Brown wants to plug in the actual non-physical description 'NP', whatever it may be. But dualists and physicalists have wildly differing ideas about what this description will end up looking like. Physicalists presumably think it will be empty (i.e. tautologous: devoid of information), whereas dualists think it will list all the qualia facts Q. Let's consider these two possibilities in turn.

• Suppose NP states nothing (i.e. is tautologous). The argument is then as follows: "It's conceivable that (TAUT & ~Q). Therefore dualism is false." Well, that's clearly invalid. The possibility of ~Q is clearly compatible with dualism!


• Suppose NP just states Q. Then the argument is: "It's conceivable that (Q & ~Q). Therefore dualism is false." But in this case the first premise is transparently false.

So, you see, whichever way we fill out the precise details of the parody argument, it very clearly fails (unlike the original zombie argument which, though controversial, can at least get off the ground, since "P & ~Q" is neither trivial nor trivially false).

You might wonder: if the parody is so atrocious, why would anyone have ever been tempted to take it seriously? I think the answer is just that they haven't fully appreciated how restrictive the form of conceivability inference relied upon by the original zombie argument is. If one assumed that the zombie argument must take the sloppy (opaque) form introduced earlier, then one may well be right to see the parody as being of the same form, and hence undermining that form of argument. I'm happy to grant this: opaque conceivability arguments are no damn good. But - once again - this does nothing to impugn the zombie argument, since it can be stated in a more transparent form (as described*).

* = (Of course, I've really only provided a schema of the argument. The real thing will have the placeholder 'P' replaced by a long, transparent description. But insofar as we have a rough grasp of what P will look like, we can draw tentative conclusions about whether the expanded version of 'P & ~Q' is likely to be conceivable.)

Compare Brown's recent response:

The zoombie world is a COMPLETE non-structural/non-functional duplicate of our world. NP is not just some random list of non-physical properties! It is a complete list of the actual non-physical properties. So, if there are no non-physical qualitative properties in NP then the actual qualitative properties that you and I enjoy are not non-physical properties.

I hope it's clear by now that this is a wildly different form of argument from the zombie argument I've given. Instead of using NP as a mere placeholder for some list of non-physical properties (neutrally and transparently specified), he is explicitly building in its status as "a complete list of the actual non-physical properties". So when he asserts that (NP & ~Q) is conceivable, this is mere shorthand for the same old sloppy, opaque conceivability claim as before (viz., that conceivably, "a world could have all the actual non-physical properties but lack qualia"). In effect, he is (unwittingly) pretending to give a precise, transparent specification by mimicking the superficial form of my argument, when in fact the underlying form and actual content of his argument hasn't changed a bit from the sloppy version.

To recap: a legitimate conceivability argument should begin by providing a transparent, uncontested specification of what is to be conceived. The zombie argument does this, by taking whatever specification 'P' the physicalist likes (which the dualist will not contest). The 'zoombie' parody, by contrast, does not meet this condition. This is indicative of further problems downstream -- and indeed, when we consider the two main candidates for fleshing out 'NP' (namely: TAUT or Q), we immediately see that neither has the faintest hope of grounding a conceivability argument against the dualist. The entire force of the parody argument rests on the opacity of 'NP', and -- unlike the real zombie argument -- it cannot survive being rendered transparent.

[This is all just to elucidate the objection originally stated in a couple of sentences in my 'zombie review'.]

Parfit's Triviality Objection

Meta-ethical naturalists hold that normative claims, e.g. 'You ought to phi', state purely natural facts, e.g. phi-ing would maximize happiness. Parfit objects that, unless we distinguish right-making features from rightness itself, our first-order normative ethical theories will be rendered trivial.

Suppose we're utilitarians. So we accept the following principle:

(U) an act is right just in case it would maximize happiness.

A non-naturalist utilitarian may make the substantive claim that whenever an act has the natural property of maximizing happiness, it thereby has the further (irreducibly normative) property of being right. If a naturalist collapses these, then their assertion of (U) seems less substantive. They are simply asserting the fact that whenever an act has the natural property of maximizing happiness, it has this very property. But that's trivial.

A complication: examples like (E) 'molecular kinetic energy is the same as heat' show us that metaphysically 'trivial' identity statements can nonetheless be informative or cognitively significant. This may be so if, although both sides of the identity statement refer to one and the same property, they refer to it in different ways (or by way of different reference-fixing descriptions). For then, in addition to the explicitly stated trivial fact, we may learn an implicitly stated substantive fact. Parfit explains:

This claim [E] gives us important information because the concept of heat is the concept of the property that is related in certain ways to certain other, different properties. (E) can be restated here as:

(F) having molecular kinetic energy is the property that can make an object have the different properties of being able to melt solids, turn liquids into gases, cause certain sensations, etc.

Here is the challenge for the metaethical naturalist: they must explain what substantive natural fact is implicitly conveyed by the likes of (U), to make such moral claims informative.

Parfit doesn't think this challenge can be met. But why not? I would have thought the obvious response for a naturalist is to adopt a kind of moral functionalism (a la Frank Jackson). The reference of 'ought' is fixed by its conceptual role R in our cognitive economy. Then, when the naturalist utilitarian claims that we ought to maximize happiness, they are implicitly conveying the substantive fact that the property of maximizing happiness has the (further) property of being such as to fulfill role R.

The problem is made vivid in Parfit's example of the naturalist utilitarian who horrifies the hospital Ethics Committee by endorsing the surreptitious murder of a patient for purpose of stealing their organs and saving more lives on net. Parfit imagines the naturalist responding as follows:

When I claimed that I ought to kill this patient, I was claiming only that this act would maximize happiness. I was not claiming that this act would have the different property of being what I ought to do. On my view, there is no such different property. Being an act that would maximize happiness is the same as being what we ought to do. Since I was claiming only that killing this patient would maximize happiness, no one has any reason to reject my claim.

But this does not seem to capture how naturalists tend to understand their own view. (The imagined 'distancing' of oneself from the ordinary understanding of moral commitments seems especially inapposite.) When naturalists make first-order claims about which natural features are right-making features, they are not only claiming that these features are self-identical (though that may be the explicit content or truthmaker for their claim, just as when scientists claim that heat is molecular energy). Rather, they are conveying the additional, important information that this feature is what the rest of us are ultimately referring to with our ordinary moral talk. That is: maximizing happiness (or whatever) is the natural property that is ultimately picked out by the reference-fixing descriptions R associated with the normative terms 'right', 'ought', etc.

There are other objections that could be made at this point, e.g. concerning whether the naturalist can really capture the full normative force of these claims. But I think the triviality objection fails.

Update: It turns out I was a bit quick in dismissing this argument.  Parfit appeals to the premise that a claim like (U) is not merely informative, but more specifically that it is a positive substantive normative claim, which attributes a distinct normative property to acts that have the property of maximizing happiness (or whatever).  It's much less clear whether the naturalist can meet this demand, since the "further information" they take us to be implicitly attributing seems to be more sociological or linguistic in nature, rather than attributing any further normative property.

Culpable Ignorance and Double Blame

[Significantly updated and moved-to-front from Feb 7.]

Suppose Adam wrongly - i.e. without sufficient reason - kills Bob. Though there's no real justification for killing Bob, such an action may nonetheless be excused if it was done unwittingly, at least if the agent's ignorance was itself non-culpable. But what if Adam was previously negligent -- i.e. his ignorance culpable* -- does this mean he's blameworthy for the action after all (at least to some degree)? I think the literature on this topic has been plagued by some serious conceptual confusions. Let me explain.

Holly Smith, in her seminal (1983) paper 'Culpable Ignorance' [pdf], considers cases in which an earlier, "benighting act" foreseeably causes our agent to later be deprived of evidence that would have led him to refrain from committing the bad "unwitting act". In one example, the agent impermissibly leaves behind his driving glasses, and so unwittingly runs over a pedestrian that he otherwise would have seen. (As will become clear, I think these are deviant cases of 'culpability' in ignorance, but more on that later.)

As HS points out, all can agree that Adam is blameworthy for the benighting act. (This is meant to be built into the specification of the case.) Further, there's a clear sense in which this makes him responsible for the ultimate outcome: "the unwitting act is a risked upshot of the benighting act", and we're responsible, roughly speaking, for the foreseeable risks that eventuate from our reckless actions (compare Russian roulette). HS thus diagnoses the debate over whether culpable ignorance is exculpating as just an instance of the moral luck debate, i.e. whether agents become more blameworthy if the foreseeable risks of their reckless or negligent acts actually eventuate.

This seems a misdiagnosis.** Let's distinguish two ways in which Adam - already culpable for the benighting act - might come to be doubly blameworthy due to the occurrence of the unwitting act. It could be that the eventuating harms render him more blameworthy in his earlier act of negligence. That's 'moral luck'. Or it could be that he is now blameworthy on more occasions: not just in the benighting act, but in the unwitting act (qua act) as well. That would be 'culpable ignorance'* failing to excuse the unwitting act as a second occasion of blameworthy agency.

Note that one might think that (even culpable*) ignorance excuses the unwitting act, in the sense that we cannot blame Adam qua agent of the unwitting act, whilst nevertheless blaming him (qua agent of the benighting act) more because of this bad outcome. So the question whether culpable ignorance excuses the unwitting act is orthogonal to the 'moral luck' issue of whether the unwitting outcome makes the (earlier) agent more blameworthy.

[N.B. If you're suspicious of this talk of "earlier" and "later" agents, you can easily translate it into talk of how the later event makes the enduring agent more blameworthy "in" acting as they did earlier. That is, I'm not really making any metaphysical claims about persistence through time here. I'm just using such talk as a convenient shorthand for locating the agent's responsibility. We may speak of that exercise of agency in virtue of which one is eligible for moral assessment; in which responsibility resides; or (equivalently) to which our moral assessment attaches. Call this the focal act. The crucial point is that one may be responsible in acting, in a completely different sense from how one may be responsible for an outcome or event (or even 'for an action', considered as a mere outcome or event, as opposed to a focal act). HS runs together 'focal actions' and 'acts-as-events' in a way which causes her to seriously misconstrue the issue at hand, or so I claim.]

Another way to see this point is to note that it doesn't seem to make any difference for HS's cases who ends up performing the unwitting bad act. She is talking about blaming Adam for Bob's death in the exact same way that we would blame Adam for Bob's death if he had negligently misplaced Cid's glasses, and Cid subsequently crashed into Bob. That is, we're blaming Adam for the outcome of Bob's death, and not (in any further sense) the act of killing him. In either case, the act of Adam's to which our blame attaches - the focal act - is the reckless 'benighting' act of depriving a driver of his glasses. This is clearly to change the subject, if what we're interested in is whether ignorance excuses (blocks blame from "attaching to") the unwitting act (as a locus). On the other hand, if we're interested in whether Adam is blameworthy, in his negligence, for the event of Bob's death, any ignorance in the "later" actor -- whether Cid or Adam himself -- is just obviously irrelevant.

Simply put: we are responsible, though (or 'in') a focal action, for various outcomes. If those outcomes are bad, there's a prima facie case for thinking the agent, in acting as they did, blameworthy. We may then ask whether there are any 'excuses' or defeaters for this presumption. These are things that may excuse the agent's action, rendering them no longer liable (again, in so acting) for various consequences. One possible excuse is ignorance. But note: By this it is not meant that if the agent becomes ignorant consequently from their act, this somehow blocks the agent from liability for any further downstream consequences of their focal act. That would be daft. No, the relevant 'ignorance' (ignorance that is eligible to excuse an agent's action) must be present in the act itself -- i.e. the act must have been done "from ignorance". It is ignorance concurrent with the focal act -- the exercise of agency in virtue of which the agent is prima facie eligible for moral assessment (positive or negative depending on, amongst other things, the consequences resulting from the act).

So, if we are considering complicated examples in which one action has as its causal upshot another act's performance, we need to be clear if the former is being treated as the focal act, and the latter as a mere event. Because if we're interested in whether ignorance (e.g. of some feature or outcome) excuses the agent's action, we had better be talking about ignorance that is concurrent with the focal action, and not ignorance that is concurrent with the mere downstream event. But this is precisely what HS fails to do. Her examples all concern cases of ignorance downstream of the focal act. As such, they are not even formally eligible to serve as excuses for the agent's action. Her very question is ill-formed.

P.S. My follow-up post considers more appropriate cases, i.e. of culpable ignorance concurrent with a focal act.

* I should note that 'culpable ignorance' is being used here in a potentially misleading way. HS really means something more like "self-inflicted ignorance", or "ignorance for which one is responsible in virtue of a prior culpable act". But that's just not the same thing as ignorance that is culpable or unreasonable in itself -- as I emphasize in my follow-up post. A 'snapshot theorist' may hold, as I do, that self-inflicted but intrinsically reasonable ignorance is indeed wholly exculpatory in the relevant sense, whilst intrinsic or 'synchronically culpable' ignorance is no excuse.

** [I owe the basic 'misdiagnosis' idea to Liz. The arguments - and broader conclusions - are my own.]

Skepticism, Rationality and Default Trust

Intellectual virtue lies between the two extremes of dogmatism and radical skepticism. The irrationality of dogmatism is generally appreciated; less so the irrationality of radical skepticism (at least among my intro M&E students!). Let's attempt a remedy.

The global skeptic insists that we should question everything, trusting nothing: apparent perception, memory, "rational intuition", even methods of reasoning -- anything that can be doubted must be suspended from our minds and considered 'guilty until proven innocent'. The obvious problem with this stance - if followed assiduously - is that it entails rejecting everything, leaving one with no mind at all. This is the crucial point: just as a pure 'blank slate' cannot learn from experience, so must we rely on various assumptions if we are to learn, reason, and act rationally at all.

That's not to say we should dogmatically refuse to question our beliefs and practices. It's just that we cannot question them all at once. Any individual belief or assumption may be tested in light of the other things we (provisionally) take to be true, and revised if some incoherence is found. But we have to reason from the beliefs we have -- as Neurath famously put it, our mind is like a ship at sea, and even as we replace a faulty plank we must trust our weight to others.

The mere fact that those other beliefs can likewise be questioned does not suffice to show that we can't reasonably accept them (again, provisionally) in the meantime. The skeptic's claim to the contrary is itself a questionable belief, and not one we have any reason to accept. (It may seem intuitive at first, but once we fully understand the implications of this claim, it would be crazy to accept it -- or so I argue.)

To begin with, it's worth emphasizing that global skepticism is self-defeating, insofar as it implies that we should not trust its recommendation to trust nothing. Or, as Railton writes (in 'How to Engage Reason', Reason and Value, p.186):

Hume observed that [the sceptic] displays a touchingly unsceptical attitude towards the power of argumentation and his own powers of thought and memory. We might add: toward his own command of language and the content of this thoughts. Remove this default confidence, and he can no more declare his words to be 'giving an argument for scepticism'--an intentional action--than 'giving a recipe for haggis' or 'scat-singing without a tune'.

Hence my earlier claim that one who truly internalizes global skepticism is no longer capable of intentional thought or action at all. To accept global skepticism is to forsake any hope of rationality. It is the ultimate intellectual black hole. Railton (p.187) draws an important lesson:

Default trust, however 'blind', is not inherently blinding. On the contrary. If the sceptic trusts his ability to speak English and draw the conclusions demanded of his premisses, and I trust my own appreciation of his argument, we will both see (no longer be blind to) a problem that I have in defending my beliefs: Where I previously had hoped to be able always to have a reason for whatever I believe, taking nothing 'blindly' or 'without reason', I now realize that this hope is impossible--some things cannot, without regress or circularity, be argued for.

I trust the reader will by now agree that global skepticism is a non-starter, as any rational agent must take some things 'on trust' if they are to be capable of reasoning at all. Still, one might ask, what of slightly restricted forms of skepticism? Couldn't one consistently hold the more traditional skeptical view that it's just our sensory experiences and/or inductive practices that shouldn't be trusted? The traditional skeptic is willing to trust in reasoning as much as anybody is. They simply have different expectations about the external world (or "affirm a different prior") from the rest of us. In particular, they hold that all possible worlds are (a priori) equally probable, whereas we anti-skeptics consider some (perhaps simpler or more regular-seeming) distributions of properties across space and time to be more probable than others.

Sure, there may not be anything inconsistent about traditional skepticism, so defined. (Though such a skeptic has no reason to expect that they will continue to exist long enough to finish their thought.) But nor is there much reason to accept it, or indeed to find it any more credible than the claim that the world just came into existence 5 minutes ago. Admittedly, traditional skepticism is motivated by a premise that seems plausible at first glance, namely: "the reasonable 'default' view is to start off by assigning each possible world an equal probability of being actual." That sounds fairer and more reasonable than an a priori bias in favour of, say, worlds where memories are typically true--representing times and events that really did happen. But I suspect the intuitive plausibility of this is an instance of us being misled by overly-abstract principles. When we consider every more particular judgment or knowledge-attribution we are inclined to make, is it really plausible to think that the one abstract principle is more credible than all our conflicting particular judgments - and practices - combined? Colour me (cough) skeptical.

Upon reflection - in light of our actual beliefs - it seems we have most reason to reject the skeptic's principle. Though it seemed plausible at first, it is inconsistent with other claims that most of us find much more plausible. Here's another: rational agents should learn from experience. Skeptics -- even of the merely 'traditional' variety -- can't. As Railton writes (in 'Rational Desire and Rationality in Desire'):

It is well known that in order to learn about one's environment (whether one be human, animal, or teachable computer) it is not enough to have ample sensory input and plenty of memory registers to fill. The learner must also bring some expectations--such as expected dimensions of similarity (the "implicit quality space"). Otherwise, experience will simply accumulate in its infinite diversity, and all experiences will be equally relevant or irrelevant to one another. No lessons will be extracted.

Carnap gave an elegant demonstration of this point within the theory of logical probability. He asked us to consider a confirmation function that began (sensibly, it would seem) with no "prior bias", i.e., that assigned the same non-zero probability to every possible state of the world (what he called "state descriptions"). This function would, even given indefinitely large amounts of information about past states of the world, still assign the same probability to every logically possible way of extending this history into the future. In a fundamental sense, it could not learn from experience.

So I think it's pretty safe to conclude that (even the traditional form of) radical skepticism isn't rational. There's no guarantee that the rest of us are any better off, of course, but at least we have a chance. We do the best we can -- and we're yet to see any good reason to think that some other way is better.

One may be left feeling unsatisfied: the best we can do may not seem good enough. But since turning to skepticism is even worse, it seems we will just have to learn to live with tentatively trusting in the reliability of our perceptions, despite the uncertainty of it all. At least we can learn such things -- and that is something to be thankful for.

Against a Defense of Future Tuesday Indifference

How should we understand Parfit's example of the hedonist with future Tuesday indifference? Sharon Street ('In Defense of Future Tuesday Indifference') distinguishes two possible interpretations, but I want to urge a third.

Suppose first that the agent is only indifferent to future Tuesdays, and on the stroke of midnight his preferences change so that he regrets his earlier decision to schedule an agonizing operation for this day. Such fundamental preference changes complicate the case. As Street points out, the different temporal stages of the agent would effectively be at war with each other -- the earlier ones plotting to ensure that the Tuesday-stage suffers agony (supposing this is necessary to spare their later stages, which they care more about, from some lesser pain), and then the Tuesday-stage trying to undo this plot against himself. That is, the Tuesday-agent has a diverging deliberative standpoint, from which he won't endorse or carry out the intentions of his earlier stages. In this sense, it's almost like a new agent temporarily takes over the body each Tuesday, which raises complications regarding whether the prior, Tuesday-indifferent stages are doing something morally objectionable in imposing suffering on the Tuesday stage against "his" will.

To avoid such complications, we may instead suppose the agent is tenselessly Tuesday-indifferent. His reflective preferences remain the same, even on Tuesdays themselves. However, Street argues that when we imagine this agent in vivid detail, he is not so obviously irrational. For during his painful experience, he maintains his "meta-hedonic" indifference to the pain, and so we might think that he achieves a state of emotional calm, and so doesn't really "suffer" from his pain in the ordinary way. (Compare, e.g., a Buddhist monk distancing himself from the searing pain of hot coals.)

However, this scenario too strikes me as containing some potentially confounding complications. In particular, it's no longer clear that the Tuesday experiences we've described are really as painful (hedonically bad) as the experiences felt on other days, and so rather than an agent whose preferences make arbitrary references to Tuesday as such, it seems we've instead described an agent with the more ordinary preference to experiences "pains" on days when they won't cause him suffering (and given his odd constitution, this happens to be Tuesdays).

To clarify this, we need to distinguish two further versions of (tenseless) Tuesday-indifference. Street offers a non-conceptual interpretation, whereby the agent ('Indy') is simply constituted (perhaps due to some bizarre evolutionary story) such that he undergoes regular cycles of psychological transformation. In particular, Indy feels a Buddhist-like 'detachment' from any pain inflicted during special periods (that happen to coincide with Tuesdays), and this indifference naturally carries over to his prospective and retrospective evaluations of "pain" experienced during the special period.

Note that Indy's pain-indifference is prior to his beliefs about what day it is. Locked in a dungeon and deprived of any other temporal cues, he might one day notice a psychological change in himself ("I no longer care about present pain experiences"), and thereby infer that it's Tuesday. This clearly isn't the kind of agent we normally have in mind when talking about Future Tuesday Indifference. Most importantly, Indy's changing psychology corrupts the thought experiment. His phenomenal experience of pain-on-Tuesday is qualitatively different from how he experiences pain on other days, so it could be this qualitative difference, rather than the purely temporal difference, that his preferences are tracking.

So I think it is more worthwhile to consider a psychologically uniform agent with explicitly conceptualized Tuesday-indifference. That is, we should imagine an agent whose psychology is consistent across time, and whose preferences make special reference to "Tuesday" as such (under that description). This means that his beliefs about what day it is will affect his behaviour: in particular, if he falsely believes that today isn't Tuesday, his subjective experiences of pain will feel as agonizing as they do on other days.

This strikes me as the 'pure' version of the thought experiment. After all, to assess Future Tuesday Indifference, we need to hold all else equal, and that means ensuring that the experiences in question are qualitatively identical -- not differing in any phenomenally discernible, and hence potentially hedonically relevant, respect from what he'd suffer on any other day. So let's ask the agent to choose between the following options:

(1) We will wait until Tuesday, and then hook him up to an experience machine that will give him the total phenomenal experience of intense suffering. (This may include brainwashing him into thinking it's Wednesday, or otherwise ensuring that he doesn't "dissociate" himself from the pain in any way that introduces subjective differences.)

(2) We wait until Wednesday, and merely inflict moderate suffering with the experience machine.

The genuinely Future-Tuesday Indifferent agent will choose option (1), by definition. Will Indy?

This poses a dilemma for Street. Indy previously seemed less irrational precisely because we could interpret his Future-Tuesday Indifference as largely compatible with a kind of impartial hedonism: he wasn't really suffering on Tuesdays, after all. But this new choice prises the two motivations apart. He can sanely choose to minimize his suffering, by opting for (2), but then we see that he's not really indifferent to Tuesday suffering after all. (It was just never an issue before, due to his quirky ability to ignore pain on Tuesdays.) Or he can choose option (1), but then we see that his apparent reasonableness was an illusion.

In any case, genuine Future-Tuesday Indifference seems (intuitively) as irrational as ever.

Ignoring Reality Ain't So Ideal Either

Paul is unimpressed with non-ideal theory that takes into account agents' own moral failings or unreliability when prescribing what they ought to do:

Moral obligations ought not to depend on an agent’s character. We ought to insist that an agent with a bad character act entirely like an agent with a good character...

That seems inadvisable. Cases like the bad squash loser show that we ignore our moral failings at our peril. For a more high-stakes example, suppose God offers me a deal that's a freebie for a saint, but insanely risky for the rest of us:

(SaintOrSinner) God will save one innocent life now, but if I ever fail to meet a moral obligation for the rest of my life then a million innocents will be tortured and killed.

Should I take the deal? If I'm a saint (and know it), then it's clearly obligatory: I can save a life at no real cost, since - as a saint - I'm sure to meet my later obligations whatever they may be. But since I'm not a saint, it's very clearly impermissible -- I'd almost certainly be condemning a million innocent people to torture and premature death.

More generally, it's fallacious to move from the fact that it would be good to do P and Q to it would be good to do P. Remember that evaluations of act-aggregations and of individual acts may diverge:

If my future self cannot necessarily be trusted to do the ideal thing, this could radically alter what current decision would be for the best. Suppose my currently φ-ing could lead to (i) the best possible outcome if I were to follow this up with a series of acts S which I could, but actually won't, perform; and (ii) the worst possible outcome otherwise. In those circumstances, it is not best for me to φ -- doing so would have very bad consequences, due to my subsequent failure to S -- even though it's part of the best possible life. The divergence arises because at any given time I can only choose how to act then; I cannot perform a lifetime's aggregation of acts with a single decision.

So we need to be clear about what normative question is being asked. It may be advisable both (i) to do P-and-then-Q, and yet (ii) not to do P (because one is unlikely to follow up with Q, and the consequences of P without Q would be disastrous). This is perfectly consistent, because the two answers address different questions: one concerning the aggregation of acts that would be best, and the other concerning what present individual act would be best. It's just a basic fact that in non-ideal agents, these may diverge.

Two postscripts:
1. Although we will typically be more interested in deliberating over individual acts than collections thereof (since we can really only perform the former), there may be exceptions. See, e.g., my complaints about illegitimate appeals to 'Political Reality': they may be fine from an individual perspective, but sometimes we want to deliberate from an explicitly collective perspective, about the question of what 'we' (rather than just 'I now') should do. This shift in focus can be vital for breaking out of a bad equilibrium.

2. My old post on 'Accommodating Unreason' offers some reasons why, at least in low-stakes contexts, it may be better to treat people as if they are morally more reliable than they really are.

Intellectual Black Holes

The value of freedom sometimes calls for constraint: against selling oneself into slavery, for example. Is there an intellectual analogue of this? We generally value open-mindedness, and a willingness to 'follow the arguments' wherever they may lead. But what if some ideas were so corrupting that to consider them seriously would risk undermining one's future capacity for rational thought? It would then seem that there are some ideas that we must close our minds to, if we truly value open-mindedness (and not just for the present moment). This seems a curious possibility.

Such 'intellectual black holes' might fall into either of two classes:

(1) Psychological traps depend on contingent quirks of human cognitive architecture. These might (in principle) be completely arbitrary: we can imagine a creature whose head explodes if it comes to believe that cows eat grass. More realistically, it seems conceivable that some (even true) beliefs might interact to ill effect with our evolved heuristics, biases, and emotional dispositions.

(2) Rational traps, on the other hand, are general to any rational mind. This makes them more philosophically interesting, insofar as they reveal propositions that are essentially (and not merely contingently) antithetical to rational thought.

Now for the big question: are there any such 'intellectual black holes' -- ideas which, if accepted, would undermine one's rational capacities? (And if so, does that necessarily mean we are justified in believing the "trap proposition" to instead be false?)

Perhaps the most obvious example is epistemic nihilism: the view that there is no such thing as epistemic rationality -- all beliefs and arguments are equally good (or bad), and rational persuasion is impossible. Notice that if someone came to really, truly believe this, then it would be utterly impossible to reason them out of it: they would be incapable of treating anything you said as a reason worthy of consideration. Their mind could only be 'rescued' by some non-rational intervention: brain surgery, perhaps.

What other such cases can you think of?

Personal Bias and Peer Disagreement

The epistemic issues that arise when considering how to rationally respond to evidence of one's own bias are, I think, exactly the same as those discussed in the peer disagreement literature. The core issue, in both cases, is how to weigh higher-order evidence (of our own possible unreliability) against the first-order evidence. If I believe P when an epistemic peer judges my evidence to instead support not-P, this seems isomorphic to a case where I know 50% of people in my position will be mistaken due to bias [corrected, thanks Daniel]. So the same three basic positions suggest themselves in either case:

(1) Higher-order evidence trumps. This is called the 'Equal Weight View' in the peer-disagreement case. In the personal bias case, we might call it the 'Automatic Adjustment View', since it claims you should adjust your initial judgment to compensate for merely possible bias, no matter what the first-order evidence actually recommends. This is essentially the claim that you're in the same epistemic position whether you turn out to be actually biased or not. (After all, the argument goes, from the inside you can't tell which of the two positions you're in. So any rule you can follow will have to treat both situations the same.)

(2) First-order evidence trumps. In the peer disagreement case, this is called the 'Asymmetric No Independent Weight View' -- since the thought is that whoever is actually right about what the first-order evidence supports may thereby "stick to their guns", giving no weight to the opinion of their mistaken peer, but the mistaken person should of course change their opinion to what the evidence really supports. (And too bad if they don't realize this.) This is essentially the claim that learning of possible bias doesn't change your epistemic situation; either the evidence supports P or it doesn't, and your own (in)capacities don't change this fact.

(3) The Total Evidence View. Finally, we might think one is rationally required to take all the evidence into account. Neither first- nor higher-order evidence necessarily trumps the other; though in particular cases it may do so. Note that the inclusion of the first-order evidence introduces some 'asymmetry', but even the correct person may be required to compromise their initial judgment somewhat in light of the (misleading, as it happens) higher-order evidence of their unreliability.

Note that it would seem absolutely bizarre for someone to hold different views about peer disagreement and personal bias. If you think that first-order evidence trumps in case of peer disagreement, it would be terribly inconsistent to suddenly turn around and say the higher-order evidence trumps in case of personal bias. The two issues go hand-in-hand. I'll wrap up this post by noting two further points of correspondence:

(A) The 'equal weight' and 'automatic adjustment' views both gain their plausibility from considering perceptual (more precisely: non-inferential) examples. But this is misleading, because we're effectively imagining a case where there is no (other) first-order evidence; the judgment itself is all we have to go on, so of course unopposed higher-order evidence about the reliability of this judgment is going to be decisive.

(B) Any view which disregards the epistemic importance of first-order evidence is decisively refuted by the problem of implausibly easy bootstrapping. Here I adapt Tom Kelly's objection to the EWV to instead apply to Erica Roedder's automatic adjustment view:

Suppose I receive an incoherent C- paper, and irrationally misjudge it to deserve an A. If the author happens to be a black student and I have evidence that instructors are typically biased against black students, does that automatically mean that all things considered I should boost the grade and give it an A+? Surely not. My initial judgment may be some evidence that the paper warrants an A+. But it's not the only relevant fact. In particular, it can't make up for the fact that the details of the paper itself constitute overwhelming evidence that the paper should be given a C-. The presence of higher-order evidence doesn't excuse me from concluding what the first-order evidence demands.

In short, the AAV (if understood as making positive claims about what I rationally ought to judge) makes it too easy for me to get things right. No matter what loony judgment I initially make, if I then amend it slightly in response to higher-order evidence of bias, then my resulting judgment is guaranteed to be rationally justified! Too easy.

P.S. For background, see Tom Kelly's excellent paper: 'Peer Disagreement and Higher Order Evidence' [pdf].