This competition entry has been selected for publication by the Forum team.
Entry for the Cluelessness Critiques Competition. Route: Option 1, challenging a premise.
Target: P1, the normative premise.
A summary of DiGiovanni's argument:
We "c-prefer A over B if the reason we prefer A is an impartial altruistic comparison of the actions' possible consequences".
P1, the normative premise, says that to justify c-preferring A, "it's not enough to say (e.g.) that A seems heuristically good". We need to argue that A has higher expected value "broadly speaking, meaning: In some sense we 'expect' that, if we were idealized agents who could aggregate all of A's and B's possible consequences into literal EVs, then we'd say A has higher EV". We ourselves don't need literal EVs. Otherwise "it's unacceptably arbitrary to c-prefer A".
P2, the conceptual premise, says that if our understanding of those consequences is "sufficiently coarse-grained", we have no argument for expecting our idealised self's EV for A to be higher, lower or equal to B's, so the two "EVs" are incomparable. Its sub-premises rule out precise EVs (a) and forced best guesses (b).
P3, the empirical premise, says our understanding is indeed that coarse-grained, whether we model the "EVs" formally (a) or argue informally (b).
The conclusion: "We aren't justified in c-preferring any action over any other".
I'll show that the argument depends on P1, and that P1 presents a false dichotomy. I don't dispute P2 or P3. P3 is a claim about our epistemic position, which I grant in full. My objection is to what P1 counts, not to how much we can see.
The false dichotomy is in the word "all". P1's standard is what an idealised agent would say having aggregated all of an action's possible consequences. What he relaxes is our access to that standard, not the standard itself. So either you track the total of everything an action might bring about, or your preference is arbitrary.
Not every consequence is a candidate to justify an impartial preference.
If a consequence bears on how a being fares, that grounds a claim. If it bears only on whether a merely possible being comes to exist, it is impersonal value. It may still be good to bring about, but we have no duty to do so. This follows the procreation asymmetry: we are "in favour of making people happy" but "neutral about making happy people" (Narveson 1973: 80; Frick 2020).
The principle I defend, Cross-Category Irrelevance (CCI), says these are different kinds of consideration, and that no quantity of impersonal value weighs against a claim. Claims still aggregate with claims. What the view excludes is the aggregation of impersonal value against claims. And it isn't arbitrary, because the line between what is owed and what isn't is fixed by an independent account of reasons, not drawn to suit the case.
What this leaves is a single duty: do not impose foreseeable catastrophic harm on whoever will occupy the future. What that duty needs is not the sign of the cosmic sum, but evidence of direct harm to those who will hold claims. This remains action-guiding even while admitting some degree of cluelessness.
CCI supplies what DiGiovanni says a view of this shape must supply and hasn't found. Objecting to Clifton (2025), he writes that "at least without an argument that some decomposition of the effects is normatively privileged over others, Option 3 won't give us much action guidance" (Post 4, §4.2). CCI attempts to provide that argument, and it is a bracketing keyed to claims that stays longtermist. The components are established ideas. So far as I can find, their assembly into an answer to cluelessness is new, though Michael St Jules has referred to the analogy with limited aggregation in discussion. I don't claim to refute the argument on its own aggregative terms. I claim its target is narrower than DiGiovanni suggests, and that what survives it is impartial, longtermist, and action-guiding.
One clarification about the kind of claim this is. It is first-order and truth-apt: I say P1 is false, not that we should hedge across views that disagree about it. So this is not an instance of DiGiovanni's own metanormative bracketing, which brackets whole normative views rather than consequences.
Section 1 shows the argument stands or falls with P1. Section 2 builds the third option. Section 3 is the crux. Section 4 states how much of the cluelessness survives, and Section 5 answers objections.
P1 sets the standard at the total cosmic sum of value: what an idealised aggregator of all of an action's possible consequences would say. Once that aggregate include the vast populations that some futures contain and others lack, digital minds by the trillion and whatever those minds create in turn, P3 shows we are profoundly clueless about the consequences of our actions. Narrow what the standard counts, and P3 stops reaching it.
DiGiovanni defends P1 with a template. To every proposed non-aggregative reason he replies: even so, that doesn't tell us whether we have an impartial-altruistic reason to prefer A over B. The template sorts candidate reasons into two kinds. Either your reason isn't a comparison of consequences at all, like a rule against dishonesty, in which case it survives but falls outside the conclusion's scope. Or it is a comparison of consequences meant to be impartial and altruistic, in which case P1 says it must answer to the whole of what the actions bring about, and P3 shows it can't.
The sorting misses a third kind of reason: impartial, altruistic, and consequence-sensitive, yet answering to a principled part of the consequences rather than the total. CCI identifies that part as the consequences bearing on claim-holders' interests.
P3 concerns the precision of a quantity. P1 says that quantity is the one we need. If P1 is false, P3 can stay true of the cosmic sum and no longer constrain what we may impartially prefer. What CCI takes out of the justificatory base is the upside of that sum: the good a future would contain in virtue of containing more beings, which is owed to no one. It doesn't make the future legible in general, and Section 4 says how much stays dark. It makes a different epistemic object available: not the sign of the cosmic sum, but evidence of direct harm to those who will hold claims.
A c-preference is defined as one grounded in "an impartial altruistic comparison of the actions' possible consequences". The definition says consequences. It doesn't say all consequences. P1 adds that, and the addition assumes every consequence is equally a candidate to justify the preference. CCI is a principled reason for splitting them into claims and impersonal value, built from two existing ideas.
The first is the Procreation Asymmetry: we have strong reason to make a person's life go well, but no comparable reason to create a person merely because their life would go well. Frick (2020) explains this through conditional reasons. His principle has the form: there is reason to (if I create a new person, make it the case that this person's life is worth living). The reason takes the whole conditional in its scope, and can be satisfied by non-creation. Frick distinguishes it from the narrow "if I create a person, there is reason that their life is worth living", which being an 'if-then' statement, requires the creation of a person to trigger.
Wide scope is what generates the asymmetry, because a wide conditional can be met two ways. The person can exist and fare well, or they can not exist at all. Non-creation is a way of complying. Frick makes the point with a promise: there can be no promissory reason against declining to make a promise, even one you could keep.
I take the scope structure and apply it to freedom from grave harm, not only to Frick's threshold of a life worth living. So a claim arises where an option leaves someone in a bad state, because that violates the conditional. No claim arises where an option merely fails to create someone, because non-creation satisfies it. The boundary is modal, depending on whether a being exists in the outcome or merely could have. It doesn't turn on species, time, distance, or any other factor that would undermine impartiality.
Merely possible beings' welfare could still be good to bring about. But nothing about them is owed, so they ground no claim, and it is claims that I argue should be action-guiding.
Parfit's (1984) non-identity problem is a concern, and the responses I use were formulated by others. A damaging climate policy changes who gets conceived, so the people alive a century later owe their existence to it, and are arguably not worse off for it.
Hare (2007) and Kumar (2009) provide the response that our duty is owed de dicto, to whoever ends up occupying the future, whatever their identity. Harm is non-comparative, following Shiffrin (1999): to harm someone is to put them in a bad state, not to make them worse off than they would otherwise have been. An energy policy that leaves whoever inhabits the future in a seriously bad condition wrongs them, even though no particular person is worse off under the alternative.
The duty rests only on there being someone downstream of the choice, and on either energy policy there will be. Who those people are, and how many, does vary with the policy. A supposed duty to select the policy that results in a larger or happier population rests on the part that varies, so it falls on the impersonal-value side. Instead, our duty is to avoid causing harm to whoever will exist.
The second idea is limited aggregation (Voorhoeve 2014; Tomlin 2017). Some considerations aren't merely outweighed by stronger ones but excluded from weighing against them, so that no number of headaches outweighs a death. Standard versions draw the line by strength, between weaker and stronger claims. I draw it one boundary earlier, between claims and non-claims.
Cross-Category Irrelevance (CCI). A claim is a consideration grounded in a being's interest in how it fares, not in whether it exists. Impersonal value, the welfare a merely possible being would enjoy if created, is not the kind of consideration that aggregates against claims. However much of it is at stake, we have no duty to produce it. The categories apply to considerations, not to beings. Whether to create a being is a question of impersonal value. Harming a being once it exists violates a claim.
Nothing here gives a being a claim in favour of being created. In an outcome where it doesn't exist, the conditional reason is satisfied, so there is nothing to press. Once created, it would hold claims in full against anyone who then harmed it. That's why the suffering of beings a bad future contains is a claim, explored in Section 4.
Claims are assessed ex ante, so a claim's strength goes with the probability of the stake as well as its size. This is relevant for objection (a) in Section 5.
Inside the claims category, aggregation proceeds as usual. Claims of comparable strength add up, which is why a harm spread thinly across billions of future people can be grave. CCI is categorical across the claim boundary and doesn't apply within it.
Drawing the line here also avoids repairs standard limited aggregation needs. On those views, whether a weak claim is relevant depends on the strongest claim it faces, and that relativity is hard to stabilise across choices with more than two options (Tomlin 2017; Mann 2021, 2022). CCI's exclusion isn't strength-relative. Impersonal value isn't a competitor at all, however many the options.
CCI isn't a discount or a ceiling. It sets no upper bound on impersonal value and applies no fractional weight to it. It denies that impersonal value competes at all with claims.
CCI doesn't owe a complete ordering. DiGiovanni has already licensed principled incompleteness, holding that "money pump arguments for completeness make the implausible assumption that, when you consider A incomparable to B, you can't simply pick the option that avoids the money pump" (Post 2, §2.4, A5).
Most importantly, CCI isn't an aggregation over merely possible welfare, which is where DiGiovanni sets his scope. The arguments "apply just as well to any alternative to EV that still explicitly aggregates over possible worlds", the core problem being "the severe imprecision of our evaluations of outcomes" (Post 4, §4.2). CCI doesn't carry the impersonal remainder in the sum at an imprecise value. It treats that component as non-justificatory and leaves it out. It denies that impersonal value competes at all, so the imprecision has nothing to attach to.
Consequentialising doesn't restore the problem. Any view can be redescribed as maximising something (Greaves 2016), but the redescribed claims view maximises over claim-relevant outcomes only, and the severe imprecision lived in the impersonal value component.
Suppose A and B differ in two ways. A spares whoever will live a grave and foreseeable harm. B imposes that harm on whoever lives under it, but contains more impersonal value, because it brings into existence a much larger number of beings with good lives who would otherwise never have existed.
The case I care about is a pair of energy policies. A is strong mitigation, which forgoes a surplus. B allows unchecked emissions and spends that surplus on creating new minds. I c-prefer A.
That preference compares the actions' consequences, so it's a c-preference. It's impartial and altruistic, because it counts the harm to every claim-holder alike, whenever they live. It isn't grounded in a higher total, because it sets the additional beings aside instead of adding them in. And it isn't arbitrary, for the three reasons Section 3 gives. That's the combination P1 says can't exist. Whether we can tell that A in fact spares the harm is objection (b).
P1 has two parts. There is a standard of correctness, that an idealised aggregator of all of an action's possible consequences would rank A ahead. And there is a rationale, that "otherwise, it's unacceptably arbitrary to c-prefer A". Nothing below depends on who does the aggregating, so P2 is untouched. My objection is to what the standard counts.
There are two demands that need separating.
The first is an eligibility demand. The ground of your preference must be principled: fixed in advance by an account of what matters that you would apply to any case, rather than selected because it delivers the verdict you wanted in this one.
The second is a weighing demand. Given the considerations that are eligible, how they net out must not be an arbitrary precisification of something we can't see.
Both are legitimate, and they come apart. The eligibility demand is what makes P1 look like a standard of justification rather than a theory of the good, and it is what DiGiovanni's objection to Clifton presses: no decomposition of the effects is normatively privileged. He has put the point directly, replying in the announcement thread to the suggestion that a person-affecting framework might help: existing bracketing approaches are either "not 'impartial' in the sense that some moral patients are bracketed out for not well-motivated reasons, or not action-guiding". That is the demand CCI has to meet, and the three reasons below are my answer to it. The weighing demand is answerable domain by domain, and Section 4 answers it using DiGiovanni's own test.
Getting from eligibility to the aggregation standard needs a further claim: that a total is the only principled ground available. This implies a first-order theory of the good, holding that what matters impartially is welfare summed across all bearers, actual and merely possible. That is totalism, and it is what the claims view denies.
So P1 faces a dilemma. Read it as the eligibility demand and it is true, but the claims view satisfies it, and the premise rules nothing out. Read it as the totalist standard and it does rule out the claims view, but the rationale no longer supports it. What excludes the claims view is then a theory of the good, and P1 offers no argument for that theory beyond the charge of arbitrariness, which I have addressed.
The line is prior. It was drawn for problems raised long before unawareness: the Procreation Asymmetry, and the super-beneficiary of Shulman and Bostrom (2021), "a being that is superhumanly efficient at deriving well-being from resources", since unrestricted aggregation over possible beings generates a standing obligation to create ever more of them. A theory that requires that was in difficulty already, so the restriction isn't an escape hatch built for cluelessness.
The line is fixed by an independent account rather than by our evidence. Where it falls is settled by Frick's (2020) analysis of the form of welfare reasons, and it would fall in the same place if our forecasting were perfect. This is why DiGiovanni's objection to Clifton (2025) doesn't transfer. Clifton's criterion is explicitly epistemic: set to zero the weight of "all the reasons for which we have qualitatively less principled ways of setting weights", because our near-term weights are "much more grounded in principles we endorse". So his division moves as our evidence moves, and rival carvings are always available. That is what makes the demand for a "normatively privileged" decomposition relevant. A normative division doesn't face this issue. I am relocating his boundary, in the same framing of obligations rather than heuristics.
The line is costly. It concedes that a gentle, consensual human extinction wrongs no one (objection (d)), gives us no credit for the vast good the future might contain, and leaves much of the cluelessness standing, as Section 4 concedes. A boundary drawn to suit the case would not take those three hits.
That leaves one route from the eligibility demand to the totalist standard. Perhaps aggregation follows not from non-arbitrariness but from impartiality. DiGiovanni's text suggests he thinks so. Cataloguing what survives cluelessness, he writes that once "your reasons for choice based on the impartial good have gone as far as they can, you might choose among the remaining options based on non-consequentialist or parochially consequentialist reasons" (Post 2, §2.4, A6). The survivors are placed outside the impartial good, and an impartial reason that isn't aggregative has no cell of its own. The same equation surfaces where he says, of the strangest hypotheses on his list, that "if we're truly impartial, we can't dismiss these 'galaxy-brained' considerations out of hand" (Post 2, §2.3.1). To be truly impartial, here, is to enter every possible being's welfare into the sum.
That equation is an assumption rather than a result. Impartiality requires non-arbitrariness among those whose interests are at stake, so that no one's like interests are discounted because of who they are, which group they belong to, or when they live. The claims view satisfies that. Its only distinctions are modal, between beings who exist in an outcome and beings who merely could, and categorical, between a claim and impersonal value. Neither attends to species, nationality, proximity, or date. A future person's claim counts the same as a present person's, and an artificial person's the same as a biological one's.
One reply is that the modal line is itself a partiality: by counting those who exist and setting aside those who merely could, the view privileges one class over another. This mistakes the structure of the view. Partiality, in the sense impartiality rules out, is discounting the like interests of a being who has interests. Where no being exists, there are no interests to discount. A view that attempts to consider all non-existent beings as equal to those who exist would run into severe difficulty.
So to return to DiGiovanni's template: even so, your reason isn't yet an impartial-altruistic reason to prefer A. But on the claims view, sparing a future person from grave harm is altruistic in content, because it concerns another's welfare, and impartial in form, because it privileges no one by identity, group, or date. To insist that it still doesn't count, because it doesn't track the cosmic total, is to restate P1 rather than defend it. Nor is this a Moore shift: the work is done by an independently motivated account of justification, which replaces P1 rather than defying it.
On the aggregative view, a longtermist reason must take the form of a claim about net cosmic expected value, and P3 applies in full. On the claims view, the central longtermist reason is the negative duty stated in Section 0, read de dicto: do not impose foreseeable catastrophic harm on whoever will occupy the future.
How much of the cluelessness does CCI clear? Not all of it, but enough to be action-guiding. Where the imprecision concerns how much good a vast future would contain, how many flourishing lives fill the galaxies, CCI removes it from the justificatory base. Where it concerns how badly off the beings who occupy an outcome would be, it does not. One of his own questions is "what's the relative value of a world taken over by different kinds of misaligned AIs or humans?" (Post 2, §2.3.1). That is a question about populated outcomes, and the beings in them hold claims. So do the victims of a malevolent lock-in. Our imprecision about all of them survives CCI untouched.
So CCI doesn't dissolve outcome robustness. It makes it one-sided. The good a future might contain drops out of the justificatory base. The harm it might contain does not. That asymmetry is enough for the duty I defend, because a negative duty needs only the sign of a claim-difference on a mechanism we understand, not the magnitude of a cosmic total.
Three things keep the claim-based comparison determinate where it is meant to guide action. The comparison is of a harm to whoever occupies the outcome, so it doesn't require identifying who they are. It needs only the sign of the difference, not a magnitude, so it survives imprecision that would defeat a ranking. And because claims are assessed ex ante, a strong claim requires a high probability of a serious stake, which is what a well-understood mechanism supplies and a speculative pathway does not. Where the sign isn't legible, because the mechanism is novel and the harm ranges straddle zero, the duty is silent. I take that to be the honest boundary of the view rather than a failure.
Climate change is the paradigm of a legible case. We have mechanistic science of radiative forcing, sea-level rise and agricultural collapse, and we know, in the ordinary evidenced sense, that catastrophic destabilisation severely harms whoever lives through it. DiGiovanni's reply to the claim that some things are obviously good is that "c-preferability is (arguably) not something we can directly perceive. Rather, it is constituted by weighing up possible consequences" (Summary). But the climate duty doesn't rest on perceiving that mitigation is c-preferable all things considered. It rests on evidence that catastrophic warming inflicts grave harm on people who will hold claims.
Catastrophic harm from misaligned AI is becoming legible in the same way, and the same rule sorts it. Claims are assessed ex ante, so strength goes with the size of the stake as well as its probability, and the stake here is the killing and injuring of people who will hold claims, not the loss of a future that is owed to no one. A stake of that size grounds a strong claim well short of certainty. What it does not license is a lower evidential bar because the stakes are high. The claim still has to rest on mechanisms we understand, which is his condition (A), and an intervention still has to show that it lowers the risk rather than raising it through the unfamiliar pathways he documents.
Choosing among interventions is more exposed to cluelessness than either prohibition, so it is worth being exact about what the duty delivers. It rules an option out: unchecked emissions impose a foreseeable grave harm on whoever will occupy the future, so they are impermissible. Because that option is approximately the current trajectory, complying requires action, which is why a prohibition is not idle here even though it is not an optimisation requirement. What the duty does not do is rank the permissible alternatives, since a given policy's net effect on claim-holders, through transition costs and displacement, is harder to foresee than the harm the prohibition addresses. But it needs only one legible comparison, not a ranking. Evidence that some available alternative spares whoever occupies the future the harm the status quo imposes. The view is domain-sensitive, and it is his own condition (A) doing the sorting rather than anything I have added.
DiGiovanni's own criterion confirms the point. Unawareness isn't a serious problem, he says, where "we have (A) a deep understanding of the mechanisms determining a strategy's consequences on some scale, and (B) evidence of consistent success in similar contexts" (Post 2, §2.2). What matters "isn't the time horizon per se, but the number and familiarity of distinct pathways that could unfold", and the physics of a destabilised climate is understood as well as any pathway on his list. The test is conjunctive, and condition (B) has no purchase here, because it tests a track record of successful optimisation. A negative duty isn't an attempt to optimise. There is no record of past attempts to beat, only a mechanism to understand. What would put the duty back in the clueless zone is the insistence that it be redescribed as cosmic optimisation, and that insistence is P1's, not P3's.
(a) "You've changed the subject, or abandoned cause-prioritisation." DiGiovanni is careful that the argument concerns promotion of the impartial good, and that norms like honesty survive it (Summary). The climate duty isn't one of those content-neutral survivors. It still discriminates among ends: the prohibition on imposing foreseeable catastrophe takes priority, while the optimisation of merely possible populations drops out. His prediction that a P1-denier should end up prioritising "neartermist causes" (Summary) holds for a bracketing keyed to what we can measure. It fails for one keyed to claims, because future people hold claims in full.
Curran (2025) supports the prediction from within anti-aggregation itself. A longtermist intervention offers each future person only a tiny chance of benefit, so their ex ante claims to assistance discount to almost nothing. Her concession runs the other way: "ex-post anti-aggregative moral theories can display a preference for some long-term interventions, but this is only when the intervention in question is such that you reasonably expect to prevent a comparable harm". Since the view I defend is ex ante, I can't simply borrow that. What I can say is that her discount is driven by the tiny probability that a speculative intervention benefits any given person, and the prohibition I defend is not like that. Conditional on unchecked emissions, the probability that whoever occupies the future suffers grave harm is high, not tiny, and it is grounded in the mechanisms his own condition (A) asks for. An ex ante claims view therefore distinguishes the climate prohibition from speculative long-shot interventions without needing a foreseeability cut-off.
The verdict is a deontic longtermism, but not Greaves and MacAskill's (2025), on which the far-future stakes are so high that any plausible theory ends up "becoming more consequentialist in output as the axiological stakes get higher". CCI denies that ratchet, which is why cluelessness about the cosmic stakes can't touch it.
(b) "To know you avoid harm, you still need the impact comparison." This is the most serious objection. DiGiovanni shows that the conventional longtermist recipe, identify a high-stakes outcome we want to avoid, research it, intervene to prevent it, "isn't robust to unawareness" (Post 3, §3.3.1): given reasonable ranges of parameter estimates, "the intervention is positive under some estimates and negative under others, and it's arbitrary how we weigh up their plausibility" (§3.3.2).
My reply distinguishes optimising from constraining. The recipe seeks an intervention that is net-positive over the cosmic sum. The negative duty asks a weaker question. Is there foreseeable, direct, serious harm to claim-holders on a mechanism we understand? That is a coarse judgement of a kind he grants we can make, since "you know the Eiffel Tower is taller than the Louvre, without knowing their exact heights" (Post 2, §2.1).
A deeper version runs from Greaves (2016) through Mogensen (2021) to the paralysis argument of MacAskill and Mogensen (2021). Almost any act changes who is conceived. If a constraint against doing harm attaches to every downstream ripple, a view with constraints tells us to do as little as possible. This is a standing problem for any view that pairs constraints with an asymmetry between harming and benefiting, and it was one well before unawareness was raised. Denying P1 doesn't create it, and the claims view inherits it on the same terms as its neighbours.
The reply I take is that a constraint needs a harm it can pick out. It doesn't need a named bearer, and the de dicto reading already supplies one: whoever occupies the future is the bearer, which is what the non-identity response in Section 2 established. What ripple-space fails to supply is any description of a harm. This isn't a probability threshold, and it isn't a claim about the length or opacity of the causal chain. Unforeseeable identity-ripples yield no claim-based verdict either way. The duty is silent about them, which is not permission, and silence is no licence to treat them as zero. The view is incomplete over ripple-space, and the incompleteness licence already covers that.
Nor is the prohibition dislodged by imagining hidden mechanisms on which severe warming turns out well. He rejects that move himself, if only for an everyday prediction, where "it would be arbitrary to posit hidden mechanisms that imply drastic deviations from this prediction" (Post 2, §2.2).
The defence is domain-sensitive, as Section 4 conceded. The choice among climate interventions is more exposed than the prohibition, since a policy's net effect on claim-holders, through transition costs and displacement, is harder to foresee. AI safety isn't excluded but held to a higher bar. Recast as preventing catastrophic harm to whoever will exist, it has the climate duty's structure, but it must also show that the intervention lowers that risk rather than raising it through the unfamiliar pathways DiGiovanni documents. This is his own condition (A) doing the sorting.
(c) "This is a rival ethical theory, not a refutation of a premise." A fair worry, and DiGiovanni has sharpened it: "impartial altruistic reasons" is defined as reasons drawn from a comparison of consequences, so deontological reasons are excluded by definition rather than by argument. My answer is not that the claims view smuggles itself back inside that definition. It is that the definition of a c-preference and the content of P1 come apart. A c-preference is grounded in "an impartial altruistic comparison of the actions' possible consequences", and the claims view offers exactly that. P1 then adds that the comparison must answer to all of them. Section 3 is an argument that the addition is unsupported.
Nor was the sequence pitched at one theory. It is titled a challenge to impartial altruist action guidance, its conclusion denies any impartial-altruistic reason to prefer one action to another, and it frames the objective as expecting better consequences "accounting for all the hypotheses that bear significantly on our impact on all moral patients" (Post 4, his emphasis on both). Those formulations claim the genus, not one theory within it. Nor is the non-aggregative wing of the genus built for this occasion. Pummer (2023) grounds effective altruism in non-consequentialist reasons, and the live debate over his account concerns how far aggregation should run within it (Mann and Barry 2025). So P1 secures the conclusion only by equivocating between impartial altruism as such and its aggregative species.
(d) "Then a quiet extinction would be no tragedy, which is absurd." Suppose everyone alive freely chose to have no more children, and humanity ended painlessly. On Ord's (2020) view this forecloses our entire potential and ranks among the worst outcomes there could be. On the claims view, no one is wronged, because no one is owed their own existence: in an outcome where a being doesn't exist, the conditional reason is satisfied. The view can still grant that something of vast impersonal worth is lost. What it denies is that the loss is an injustice. And every realistic extinction, arriving by catastrophe rather than consent, kills and injures existing and near-future claim-holders, so the negative duty condemns it in the strongest available terms. The gentle case isolates the Asymmetry, and conceding it costs the view none of its action guidance.
DiGiovanni's defence of P1 allows only two kinds of reason: reasons that aren't about consequences, which survive but guide no altruism, and reasons that answer to all consequences, and so to everything we are unaware of. There is a third kind. A reason can be impartial, altruistic, and consequence-sensitive while answering to the part of the consequences that is owed to someone. P1's rationale is non-arbitrariness, and the claims view isn't arbitrary. P1's standard is total aggregation, and that is a theory of the good rather than a standard of justification. The premise fails on either reading.
What survives isn't a retreat to honesty and kindness. It is longtermism with claims. CCI leaves much of the cluelessness standing, because how badly off the occupants of a strange future would be is a question about claims. What it clears away is the upside, which is owed to no one, and that is enough for the duty that matters. Future people hold claims exactly as present people do, and the duty not to impose foreseeable catastrophic harm on them sits in the zone DiGiovanni's own criteria mark as tractable. When it comes to serious harms we could impose on people who will live, we aren't clueless.
The account of conditional reasons and Cross-Category Irrelevance draws on the author's separate work in progress; this entry is previously unpublished and self-contained.
