This competition entry has been selected for publication by the Forum team.
This paper is a constructive response to DiGiovanni's diagnosis of impartial altruistic cluelessness. It questions the presupposition of neutral evidence bases. Instead, and using an example of poor animal welfare, it highlights that before we bracket uncertainty, we must audit whether that uncertainty reflects genuine epistemic limits or is in fact representative of structured ignorance and pragmatic bias within our institutions. It argues that only once this distinction is recognised can bracketing become, not only a response to cluelessness, but a disciplined and morally responsible guide to action rather than merely a response to uncertainty.
Anthony DiGiovanni's recent sequence of papers presents one of the most serious challenges to impartial altruism. Rather than questioning the moral importance of promoting impartial decision making, especially with regards to welfare, he argues that our epistemic limitations prevent us from justifiably determining which actions actually do the ‘most good’, and therefore what we should support.
When indirect, long-term and even cosmos-wide consequences of our actions are considered; comparisons frequently become indeterminate. Assuming correct, for an impartial altruist to make decision supporting and maximising an expected positive impact loses the action-guiding evidence.
This paper adopts option 3 and offers a constructive response to Antony’s papers. I accept DiGiovanni's argument and do not challenge the central premise concerning epistemic limitations, nor do I deny that consequentialist optimisation arguably fails under interrogation of deep uncertainty. Instead, I respond to the question posed by the series itself:
What should the impartial altruist do anyway?
It is accepted that bracketing provides a promising and constructive response to decision making uncertainty. That when the total consequences of alternative actions cannot be compared, decision-makers should instead evaluate the largest set of beneficiaries (as brackets) whose outcomes remain meaningfully comparable [1.2].
This paper argues that the steps highlighted by bracketing, whilst an appropriate & important approach, are in my view, dangerously incomplete. A vital step of the argument remains absent, and hence highlight how it can be improved providing the impartial altruist with greater clarity of decision.
A central assumption made is that evidence provides a neutral basis for identifying what can and cannot be compared. Recent work on pragmatism bias in animal welfare science [3] suggests otherwise. The evidence available to decision-makers is itself shaped by systems and structural biases - funding arrangements, institutional incentives, and assumptions about what questions are practical to investigate commonly impact evidence generation. Consequently, whilst some apparent epistemic indeterminacy will reflect genuine uncertainty, other apparent areas of ignorance are produced by systematic omissions in the research and analytical process.
I therefore propose that before consequences are or can be bracketed as either a ‘point of value’ or as ‘unknowable’, an impartial altruist looking to make impact-based decisions should conduct an epistemic audit to distinguish genuine indeterminacy from structural ignorance. Bracketing remains the appropriate response to the former but not to the latter. Without acceptance of this, species egalitarianism and structural egoism will continue to dominate the animal welfare field and inappropriately influence and undermine aspects of the altruistic movement.
DiGiovanni's argument tackles the problem that EA faces: how do we choose an action or actions that produce the greatest overall good? How do we choose where the consequences of almost every significant step extending through complex causal networks can affect future generations, institutions, technological development, ecosystems, and potentially astronomical numbers of future beings. Under sufficiently severe unawareness or imprecision, these comparisons become indeterminate; they are not indeterminate by definition [4, 5,6].
This was quite wonderfully and abstractly stated by the U.S. Defence Secretary Donald Rumsfeld in 2002: "There are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don't know we don't know.
Antony correctly showed how this confusion means we very often face conundrum. Firstly, our models (of the world) become extraordinarily coarse. Second, it is certain that many relevant possibilities remain unknown or even unconceived. Under these scenarios, assigning precise expected ‘points of value’, or even forming justified best guesses, becomes increasingly difficult to the point of decision and action paralysis.
DiGiovanni argues that this is not only a practical problem but a structural feature of impartial decision-making. It is not surprising then that since actions cannot be meaningfully compared across their total consequences, the ability to provide an optimal approach fails, especially where decisions are or in some cases become inappropriately weighted (i.e. unknowns outweigh knowns) [4, 6].
The issue is “acceptable” – it is evolutionarily human to be risk averse. Hence our attempts to create precise rankings or measures under profound uncertainty risk replacing rational impartiality with either risky overconfidence or decision paralysis and no response. The potential end results are deeply troubling. If every significant action becomes incomparable, how can the impartial altruist ever guide behaviour?
Accepting DiGiovanni's diagnosis and suggested approach, however, does not need to entail decision or moral paralysis.
Humanity, in general faces, this problem increasingly. Our modern societies possess unprecedented technological power while retaining moral dispositions that evolved for small-scale communities. Human beings remain parochial, show near-term bias, are strongly influenced by identifiable victims despite having the capacity to progressively (and detrimentally) affect distant strangers, global systems, future generations and non-human animals [7,8].
Persson and Savulescu [7,8] argued that better moral motivation should drive decisions rather than flounder under epistemic uncertainty. Read alongside DiGiovanni, the two arguments reveal a double challenge:
We are both simultaneously insufficiently informed and morally imperfect (either insufficiently motivated morally or motivated egocentrically).
Neither problem is solved simply by calculating expected ‘points of value’ more carefully.
Nevertheless, for the impartial altruist moral action must remain possible and even probable because the ethical (and altruistic) approach demands avoidance of severe, unnecessary, and foreseeable suffering. Rights-based ethics, theories of justice, care ethics and many forms of consequentialism all provide strong reasons to prevent clear harms even when the total future consequences remain uncertain. In addition, the precautionary approach can be justified where risks are serious, involuntary or difficult to reverse [9].
The challenge therefore becomes identifying a decision procedure that respects our epistemic humility without abandoning our practical altruistic responsibilities.
Antony highlights that bracketing provides the most promising answer to this challenge.
Rather than insisting that decision-makers compare every consequence of every action, it is proposed that identifying the largest group of beneficiaries whose outcomes remain meaningfully comparable is appropriate. Consequences lying beyond this maximal bracket-set are provisionally set aside because the available evidence cannot permit justified comparison.
Importantly, this is not temporal discounting. Future beneficiaries remain equally important. But they are bracketed only where our understanding of how present actions affect them is understood and only removed when genuinely indeterminate.
Nor is bracketing merely a heuristic (i.e. objective measure system) for estimating an example total expected value. Instead, it makes a stronger normative claim, that unknown consequences should not automatically erase clear reasons arising from consequences that remain meaningfully comparable.
Importantly this means an altruist decision-maker can act despite radical uncertainty while accepting various information absences.
Bracketing also possesses important safeguards. The preferred top-down approach seeks the largest defensible comparison before excluding uncertain consequences. This means it avoids prematurely discarding effects that become meaningful especially when considered together where singular (1:1) comparisons can become circular.
Nevertheless, in my view the theory remains incomplete.
DiGiovanni acknowledges unresolved questions concerning sequential decision-making, information gathering and dynamic revision of bracket-sets. More fundamental, however, and in my experience important, bracketing assumes that the very evidence defining what can be compared is itself neutral.
This assumption is spurious and deserves closer examination.
Recent work by Mandel et al. [3] identifies a problem largely absent from the discussions of cluelessness. They argue that animal welfare science is affected by pragmatism bias: research agendas are systematically shaped by commercial feasibility, funding structures, regulatory priorities, practical constraints, and in my view human ego-centric amorality bordering on professional negligence.
Consequently, scientific attention focuses on incremental (health) improvements; existing production systems; easily measured outcomes usually set as an ROI; economically valuable species; and those within mass social or cultural conscious.
Questions that challenge the underlying systems, examine flourishing rather than survival, or investigate politically inconvenient alternatives frequently receive far less attention and are lost within the fog of indices suggesting erroneously that biological health = welfare (= morality = ethics).
This observation reveals a significant weakness in the current bracketing proposal.
Bracketing assumes that the boundary between comparable and incomparable consequences reflects the intrinsic limits of knowledge. In reality, that boundary often reflects institutional and individual choices about what has been, what will be, or what can be studied.
This is an important distinction and additional consideration an altruist must make.
Genuine epistemic indeterminacy arises where careful investigation genuinely fails to justify comparison. But structured ignorance, in contrast, arises where potentially relevant knowledge is absent because research questions were never asked, funding was unavailable, access was restricted, challenges were ignored, or prevailing assumptions and biases discouraged investigation.
These two situations are morally different and ethically important.
Genuine indeterminacy appropriately motivates epistemic humility and leads moral questions.
Structured ignorance generates obfuscations and wilfully blind study. The role of academia is lost – instead of being obligated to investigate and diversify evidence through academic rigour and curiosity, and to avidly avoid research omissions as evidence of moral insignificance, it falls into the Rumsfeld Cycle of “we know what we know”.
The practical implication for the altruist is that not every uncertain consequence should be bracketed. Some uncertainties represent failures of inquiry rather than unavoidable limits of our human understanding.
Before bracketing can guide actions responsibly, altruistic decision-makers should therefore examine how the evidential landscape itself has been produced or why it is indeed absent.
This paper therefore proposes a constructive extension rather than replacement of bracketing.
For the impartial altruist, the first step in any decision process should be an epistemic audit asking whether uncertainty reflects genuine indeterminacy or structured ignorance. Once this distinction has been made, bracketing can then be applied within a more robust framework.
Hence, in altruistic decision making five steps are hereby suggested:
Step 1. Undertake an epistemic audit. Distinguish genuine uncertainty from structured ignorance - only uncertainty that persists after reasonable investigation should be bracketed.
Step 2. Retain determinate reasons. Clear evidence of serious and preventable suffering must continue to generate a higher moral ‘point of value’ even when or where wider consequences remain uncertain.
Step 3. Bracketed comparisons should be extended as far as evidence responsibly permits; reductive selections must never be based on academic convenience.
Step 4. Bracketing should be continuously revised. Evidence evolves and so bracketing is always accepted as provisional. New evidence should trigger reassessment rather than permanently excluding uncertain beneficiaries; hence decision and actions must include self-evaluation criteria.
Step 5. Improve both institutions and decision-makers. Better choices require better evidence aligned with moral and ethical leadership. Better evidence requires institutions and people capable of investigating neglected questions openly and representing neglected interests especially in the face of challenge.
Taken together, these steps preserve the strengths of bracketing, address its weakness of dependence upon biased or absent knowledge, and provides progression.
The distinction between genuine uncertainty and structured ignorance can be exampled to a relevant area of current altruist discussion - decapod welfare.
There is substantial evidence that many decapods possess the capacity for pain and that procedures such as live boiling cause severe suffering [10]. The immediate welfare benefits of effective stunning before slaughter are therefore or would be expected to be well supported.
However, if we consider all bracketing, claims concerning long-term market responses, technological development, systemic legitimisation, or maybe even distant ecological effects introduce uncertainty.
Standard bracketing would recommend acting upon the comparatively well-understood welfare benefits while provisionally setting aside the genuinely indeterminate long-term consequences.
We can also highlight that decapods have historically received little welfare research, their importance within ecological trophic systems is largely unknown, that public conscious is low, and that (as a result) decisions regarding their treatment within industrial food production systems are absent compared with for example, terrestrial livestock (Mandel et al. 2026). This highlights the inherent weakness within bracketing, that uncertainty is not necessarily intrinsic unknowability, but that it reflects structured neglect reflected in industrial abuse.
For an altruist then an epistemic audit must therefore not only ask:
What do we know?
but also:
Why do we know this rather than something else?
In decapods, research may certainly consider the welfare of live boiling and even explore and compare different prior-stunning technologies, but coverage remains uneven across species, life stages and welfare dimensions [10]. Questions about positive welfare, conditions for thriving and alternatives outside existing production models are vulnerable within the broader pragmatism bias identified by Mandel et al. [3]. This means that wider unexplored questions about live storage, transport, alternative production systems, or ecological impacts where arguably greater welfare compromises occur are simply not raised.
As highlighted these omissions can be a result of structural ROI opportunities, university bias or favouritism, and technological emergence. Importantly any apparent emphasis should not (automatically) suggest others are morally neutral simply because they remain unexplored and therefore uncertain.
For the impartial altruist, the resulting decision is therefore neither simple precaution nor ordinary bracketing. Rather, it should combine definitive action where suffering is clearly evidenced; identification of the analytical gaps allowing continuing investigation of neglected harms and ongoing revision of brackets; and explicit recognition that some ignorance reflects institutional choices (and industrial favouritism) rather than unavoidable uncertainty. This latter aspect could even mark funding withdrawal from overtly biased structures.
DiGiovanni's central challenge remains compelling and highlights the need for critical thinking. To do so means impartial altruists cannot simply measure a total of expected points of value where innumerable consequences exceed the limits of justified comparison.
Bracketing represents the most promising constructive approach currently available because it allows action to proceed without pretending that radical uncertainty has disappeared, nor that future consequences do not hold value.
Yet I argue that bracketing itself requires one further step.
Before consequences are bracketed as unknowable, impartial altruists should distinguish genuine epistemic indeterminacy from ignorance (biased or not) produced by research priorities, institutional incentives, parochial and pragmatic constraints.
This distinction changes the practical meaning of uncertainty. Some consequences should indeed be bracketed because they cannot presently be compared. Others should instead motivate better inquiry.
The suggested steps preserve our epistemic humility without confusing the limits of our knowledge with the limits of our curiosity.
In answer to DiGiovanni's original question, ‘what should the impartial altruist do?’ the response proposed is and remains simple:
Act on the strongest reasons that remain after a critical examination of both the consequences of action and of the institutions that determine what is known about those consequences. Bracket genuine uncertainty but never bracket ignorance that systems and people choose to create.
