At the heart of our research is a guiding question: How can we best reduce suffering with the limited resources available to us?

This captures our central aim, but it is too broad to answer all at once. Any practical effort to make progress must clarify what kinds of suffering matter most, where and when they occur, and which actions are most likely to alleviate or prevent them.

We therefore break the question into smaller, connected parts. This helps us move from broad ethical commitments to concrete research priorities and strategic choices, while making our assumptions and uncertainties more transparent and easier to examine. In this article, we introduce a framework and show how it can help us reason systematically about where our efforts may do the most good. We hope it offers readers a clear picture of how we approach this challenge—and a useful starting point for examining the choices involved.

Philosophical foundations

Our approach is grounded in suffering-focused ethics: reducing suffering is not merely one good among many, but a central moral priority. We give particular attention to worst-case scenarios and to preventing severe suffering, wherever it occurs.

This commitment extends to all sentient beings, regardless of species, habitat, or relationship to humans. It includes humans and other animals, wild animals, and potentially sentient invertebrates and artificial beings. Where sentience is uncertain, we believe there is reason to take the possibility of suffering seriously.

We also give equal moral consideration to suffering across time. Since the long-term future may contain an exceptionally large number of sentient beings, its consequences deserve substantial attention. At the same time, we recognize the urgency of reducing suffering in the present.

Taken together, these values call for careful consideration of how our choices affect those who suffer today and how they may shape suffering over the long term.

Scenario analysis

To reason about how best to reduce suffering, we need to understand how the future might unfold. Scenario analysis involves an element of forecasting: what kinds of futures are plausible, how much suffering they might contain, and which developments we can realistically influence.

If we hope to reduce the largest amount of suffering, we need to identify scenarios that are especially concerning because of the scale of suffering they could involve. These include futures shaped by advanced AI, expansion into space, and other developments that could give rise to risks of astronomical suffering (s-risk).

Because there are so many possible futures, analyzing each one individually is impractical. Instead, we group them into categories to compare their possible outcomes. For example, suffering might arise as an unintended side effect of pursuing other goals (incidental s-risk), or be deliberately caused by agents (agential s-risk).

Several important questions remain open:

Plausibility: Which future scenarios involving suffering are most likely?
Expected outcomes: How much suffering might different classes of scenarios involve, and how might it be distributed? Is expected suffering concentrated in a small number of extreme outcomes, or spread more broadly across many futures?
Dominant risks: Which risks account for the largest share of expected future suffering, and what role do emerging technologies or expansion into space play?
Potential levers: Which developments should we pay most attention to? Which mechanisms might offer opportunities to influence these outcomes? Advanced AI and the possibility of long-lasting “lock-in” are among the areas worth examining.

These questions help direct our attention to pivotal events that could offer exceptional opportunities to reduce suffering in the long term. But the future is deeply uncertain, so the assumptions behind any scenario analysis—including which futures we consider and how plausible we judge them to be—can and should be questioned.

Risk factors and sources of suffering

A complementary approach is to examine not which futures are most likely, but how suffering could arise in the first place. By identifying and classifying its possible sources and contributing factors, we can build a clearer picture of the pathways through which suffering might occur. One useful starting point is to consider which kinds of sentient beings may be affected—for example, humans, nonhuman animals, and potentially artificial suffering in the future.

Another useful approach is to identify risk factors for s-risk: conditions that are not themselves s-risks but can make severe suffering more likely or increase its potential scale. Examples include powerful technologies without adequate safeguards, weak efforts to prevent suffering, escalating conflict and hostility, and the concentration of power in malevolent actors. Focusing on such proxies can help us identify ways to reduce suffering without having to predict every specific future scenario. Because the factors can interact, it is also important to consider how they may reinforce—or counteract—one another.

Identifying risk factors can help us work backwards from potential harms to their upstream causes, revealing earlier points where intervention may be possible. We can then assess separately whether an intervention will meaningfully change a risk factor and how much that factor affects the likelihood or severity of future suffering. This leaves an important strategic question: should we target broad risk factors (e.g., international cooperation) or narrower interventions with direct links to specific harms (e.g., animal welfare)?

Intervention analysis

Once we have mapped possible futures, sources of suffering, and the factors that could make severe suffering more likely, the next question is practical: if we pursue a particular course of action, what might change? We assess how interventions could affect future developments and, ultimately, the amount of suffering. This means looking not only at an intervention’s intended effects, but also at its indirect consequences and the possibility that it could make things worse.

We use several analytical frameworks to guide intervention analysis. One is importance, tractability, and neglectedness (ITN): For a given intervention, how much suffering is at stake? How feasible is it to make progress? And how much attention and how many resources does the area already receive?

Because forecasts are uncertain, we also ask how interventions might perform across different plausible scenarios. An action that is beneficial in many futures may be preferable to one that looks optimal only under a narrow set of assumptions. We therefore ask how to make our efforts robust across scenarios, while examining whether they could inadvertently increase suffering. This does not mean avoiding every intervention with risks: an action can have a positive expected impact even if it carries a meaningful chance of backfiring. The key is to take such risks seriously and assess them alongside the intervention’s potential benefits.

For a candidate intervention—such as work on surrogate goals, cooperation and conflict reduction, advanced AI, space governance, or moral-circle expansion (MCE)—we can examine how it might affect each of the sources of suffering outlined above. We can then rate its potential effects across those sources and combine the ratings into a single score, using explicit weights to reflect their relative importance. This score can inform our estimate of the intervention’s overall effect on reducing expected suffering.

Intervention research must also remain open to overlooked opportunities. We can ask whether any actions, cause areas, or leverage points seem unusually promising across a wide range of scenarios, yet receive little attention. Searching for such “cause X” candidates can help us avoid settling too quickly on familiar priorities—and reveal where limited resources may have the greatest effect.

What is outside our scope

To keep our research focused and action-guiding, it is also important to clarify which questions we do not aim to answer here. Some of these questions are valuable in their own right; they are simply distinct from our central task of identifying practical ways to reduce suffering.

We do not aim to develop a fundamental theory of consciousness or suffering. For our purposes, the concept of suffering is sufficiently clear to guide practical work. There are still difficult questions about which beings are conscious—including invertebrates and possible future artificial systems—and how to compare their experiences. We take these uncertainties seriously, but a purely theoretical account of consciousness is not a primary focus of this research.

Similarly, we do not focus on foundational epistemology or attempt to resolve the deeper question of whether we can reliably influence long-term outcomes or are, in effect, ‘clueless.’ The difficulty of predicting the long-term consequences of our actions under profound uncertainty is undeniably relevant to decision-making. However, given the moral urgency of reducing suffering, we focus on questions that can guide choices and interventions despite that uncertainty.

Comparing ethical frameworks and resolving abstract questions in population ethics are also outside our scope here. We take the reduction of suffering as our subject and focus on the question of how to pursue it effectively, rather than comparing suffering-focused ethics with alternative moral views or settling disputes about how to evaluate different possible worlds.

This means that we treat several principles as starting assumptions for this discussion: that reducing suffering is a central altruistic aim, that moral concern should not depend on species membership, and that the interests of future individuals deserve consideration. Different starting assumptions could lead to different conclusions, but defending these principles is outside the scope of our research on how to best reduce suffering.

Finally, we do not try to settle the complex question of when to act: whether efforts to reduce suffering are more effective today or whether it is better to wait for future opportunities. This timing question can be important, especially for large funders. Some resources, such as financial capital, can be saved for later, but research effort and many other resources often cannot simply be transferred to the future. Our research therefore focuses on the practical question of what can be done now, given the opportunities available.

Conclusion

Decomposing the question “How can we best reduce suffering?” gives us a roadmap for research and prioritisation. By turning a broad challenge into connected, more manageable questions—about possible futures, sources of suffering, and the effects of potential interventions—we can make our assumptions clearer and better identify opportunities to prevent severe suffering. This is not to suggest that every research project must follow the same sequence or method.

Some important work will instead focus on a particular area in depth. For example, we believe that research on suffering-focused AI governance and the risks associated with expansion into space are high priorities. Such research can complement the broader effort: specific fields can develop the knowledge and expertise needed, while the overarching question helps us assess how those efforts fit into the wider goal of reducing suffering.

Contents

Decomposing the Core Question: How can we best reduce suffering?