This competition entry has been selected for publication by the Forum team.
This essay will first discuss the significance of cluelessness and then argue that Anthony DiGiovanni’s proposal to assess actions by reference to global expected welfare is multiply problematic and makes an already difficult task harder. We are better advised to begin with the reasons for and against particular actions.
The significance of cluelessness
DiGiovanni is right that cluelessness presents a serious challenge to impartial altruism. We are not merely uncertain between possibilities we can describe. Our awareness set is incomplete, our representations of outcomes are coarse, and unknown unknowns may overturn our conclusions.
The problem grows as the moral circle expands. Impartial reason pushes us to recognise that similar welfare has similar value whether it belongs to ourselves now or in the future, to other people near or far, to animals or to future beings. Yet our knowledge becomes less secure as we move outwards. We understand our own circumstances better than those of unfamiliar societies, human experience better than radically different forms of animal consciousness, and near-term effects better than consequences far into the future.
Forecasting has a short horizon. We may understand an action and its immediate effect, perhaps followed by one further response. Beyond that, interaction with other systems makes prediction increasingly unstable and consequences rapidly become chaotic.
Unknown unknowns deepen the difficulty. We do not merely lack probabilities for recognised possibilities; some relevant mechanisms and outcomes never enter our models. DiGiovanni is therefore right to reject precise expected-value calculations that rest on incomplete awareness.
The problem is wider still because of our ignorance of animal sentience and welfare. We do not know which animals are conscious, the intensity and balance of their positive and negative experiences, whether many animal lives contain positive or negative welfare overall, or how radically different forms of consciousness should be compared. For example, we do not know whether habitat loss is good or bad in welfare terms because we do not know whether the animal lives lost would have been net positive. Ordinary economic, medical and environmental decisions may therefore have vast effects whose welfare significance we cannot determine.
Cluelessness should make us modest about what impartial reasoning can establish. It does not follow that impartial reasons must be grounded in a prior estimate of global expected welfare.
The problem with global welfare as a starting point
In “Cluelessness: Summary of the Argument, Why It Matters, and Counterarguments,” DiGiovanni’s first premise requires, roughly, that a justified impartial preference for one action over another be supported by reason to think that an epistemically idealised evaluator would assign it higher total expected welfare.
This places global welfare at the beginning of practical reasoning. Before regarding one action as impartially favoured, we must have justified grounds for anticipating an ideal agent’s comparison of the complete welfare consequences of the alternatives.
That standard is problematic. It requires us to anticipate a comparison encompassing all effects on all welfare-bearing beings across the relevant future, although the central lesson of cluelessness is that our evidence does not support such a comparison. Appealing to an idealised evaluator does not explain how we can form justified expectations about its verdict.
The idealisation is also unclear. Is the evaluator limited to our evidence but given unlimited reasoning power, or does it know every consequence? Does it possess a complete awareness set? Does it know which animals are conscious and whether their lives are positive or negative? Different answers produce different standards.
Nor would perfect information necessarily yield a global welfare ranking. A single welfare function requires cardinal measurement, interpersonal and cross-species comparison, aggregation across populations and a solution to population ethics. These assumptions are disputed even among utilitarians. An omniscient evaluator might know every fact about human happiness, fish pain and insect experience without discovering a common scale on which they can all be added.
Global expected welfare may sometimes summarise a successful deliberation. It cannot plausibly be the general source of every impartial reason.
A £1,000 anti-malaria donation
I suggest instead an approach based on reasons. To illustrate, consider whether I should give £1,000 to an effective anti-malaria charity.
A reasons-based analysis begins with effects for which there is some evidential basis. The opportunity cost to me of parting with the £1,000 is a clear but modest reason against donating.
The donation contributes probabilistically to preventing malaria and, in some cases, preventing a child’s death. Although the beneficiary is unidentified, the causal effect is statistically supported. The benefit to the child, together with the avoidance of grief and hardship for the family, provides a strong impartial reason in favour.
The child’s continued life will also affect future generations, economic activity, political institutions and the environment. Some effects may be positive and others negative, but their direction becomes increasingly difficult to establish.
Animal welfare presents a particularly troubling consideration. A child who survives may consume animals over many decades, and descendants may do likewise. If the animals concerned experience severe suffering, the negative effect could be very large. Yet we do not know how many additional animals the donation would cause to exist, how food systems will change, which animals are conscious, how intense their suffering is, or how their welfare should be weighed against a human life. Demographic, economic and ecological effects may favour either side but are even harder to determine.
The analysis therefore identifies a modest reason against donating, a strong reason in favour, uncertain wider human effects and a potentially very large but deeply indeterminate animal-welfare reason against. We may remain clueless about the all-things-considered answer.
That is not a failure of the method. It accurately represents what we know, what we suspect and where comparison breaks down. The reason arising from the expected benefit to the child remains visible even if its final weight cannot be determined.
Why a global welfare approach makes the problem harder
A global expected-welfare approach asks whether the complete world history resulting from the donation contains more expected welfare than the history resulting from keeping the money.
Answering requires us to model the donor’s opportunity cost, the child’s life, family effects, economic and political consequences, future conceptions, animal consumption, ecological changes and the remote future. We must then assign probabilities and translate every relevant experience into a common metric.
The additional problem is not that global welfare directs our attention to effects that reasons-based deliberation may ignore. It requires us to complete and aggregate those effects before the reasons we understand can support a preference. It therefore imposes two further demands: completion of the causal picture and commensuration of all welfare within it.
The proper order is the reverse. We should first identify the welfare-relevant reasons supported by the evidence, consider countervailing effects and uncertainty, and reach an overall judgement only where those reasons can be sufficiently weighed.
Unknown unknowns remain relevant. They may justify lower confidence, further investigation and caution. But the bare possibility of unidentified consequences does not automatically create a reason equal and opposite to an evidenced benefit.
Reasons before aggregation
A reasons-based approach resembles Jesse Clifton’s Option 3, discussed by DiGiovanni in section 4.2 of “Why Existing Approaches to Cause Prioritization Are Not Robust to Unawareness.” DiGiovanni objects that consequences can be divided into subsets in arbitrary ways and that the unknown remainder may contain reasons on both sides.
My response is that, on a reasons-based approach, the identification of significant consequences is not arbitrary. Reasons arise from supported causal and evaluative relations. The evidence that anti-malaria interventions prevent disease has an epistemic standing that an imaginable chain in which the intervention eventually makes civilisation worse does not. Well-supported indirect effects must also be considered, but mere possibility is not enough.
On unknown remainders, I agree that these may contain reasons favouring both choices. That limits the confidence and completeness of the final judgement. But it does not erase the reasons whose grounds we understand.
Conclusion
DiGiovanni is right to challenge confident global expected-value calculations. But making global welfare the necessary starting point compounds the problem. It requires speculative causal completion and controversial aggregation before any impartial preference can be justified.
We should instead begin with the reasons for and against particular actions, including indirect effects and uncertainty where the evidence supports them. Those reasons may sometimes ground a broader judgement about global welfare. At other times they will provide only partial or indeterminate guidance. Global welfare is a possible conclusion of deliberation, not the necessary source of every impartial reason.
Earlier in the text, you actually gesture at the meat-eating problem. And you implicitly build two categories: i) children saved and ii) all non-human animals (farmed, wild, etc.). And since we are clueless about overall effects on (ii), we bracket (ii) out, and anti-malaria interventions are good.
But what if instead we were dividing things this way: i) the very animals eaten by the saved children,[1] and ii) all other farmed animals + other non-human animals + the saved children. And since we are clueless about overall effects on (ii), we bracket (ii) out, and anti-malaria interventions are bad.
How is the former categorization any more legit than the latter one? (Related argument here, fwiw.)
(I like how straightforward and clear your argument is. Thanks for making it this way! Makes it easy to engage with.) :)
You can't deny we should believe these will suffer, in expectation (even if we're not certain---just like for the children that will almost but not certainly be saved and benefited, btw).