7th Oxford Workshop on Global Priorities Research

23-25 June 2021, online

Topic

Global priorities research investigates the question, ‘What should we do with our limited resources, if our goal is to do the most good?’ This question has close connections with central issues in philosophy and economics, among other fields.

This event will focus on the following two areas of global priorities research:

  1. The longtermism paradigm. Longtermism claims that, because of the potential vastness of the future of sentient life, agents aiming to do the most good should focus on improving the very long-run future, rather than on more immediate considerations. We are interested in articulating, considering arguments for and against, and exploring the implications of this longtermist thesis.
  2. General issues in cause prioritisation. We will also host talks on various cause prioritisation issues not specific to longtermism - for instance, issues in  decision theory, epistemology, game theory and optimal timing/optimal stopping theory.

These two categories of topics are elaborated in more detail in Sections 1 and 2 (respectively) of GPI’s research agenda.

Wed 23 June- Session A

TimeSession
14:30Arrival and informal networking
15:00Hilary Greaves, Welcome and overview of GPI
15:20Korbinian Rüger, “Longtermist axiology without longtermist population ethics?” CANCELLED
16:00Bridget Williams, “Age and the value of averting death in public health”
16:40Coffee break and informal networking
17:00Keynote address: Robert S. Pindyck, “Climate (and Other) Catastrophes”
18:00Informal networking

Wed 23 June- Session B
Wed 23 June- Session C
Wed 23 June- Session D
Thurs 24 June - Session A
Thurs 24 June - Session B
Thurs 24 June - Session C
Thurs 24 June - Session D
Thurs 24 June - Session E
Fri 25 June

Talk abstracts

Wednesday 23 June - Day 1 Session A

15:20-16:00 - Korbinian Rüger, “Longtermist axiology without longtermist population ethics?” - CANCELLED

Some recent approaches to population ethics are pluralist in the sense that they take into account reasons stemming from an impersonal axiology as well as reasons stemming from the complaints that individual people affected by our choices could put forward against these choices. One “problem” of such accounts is the following. While non-axiological reasons, such as individual complaints, can in principle limit the extent to which it is required or even permissible to promote the good, it is unclear how much bite they would have in practice, given that the axiological stakes in population ethics tend to be very high.  This gives rise to the following working hypothesis that my project investigates. The  implications of a plausible pluralist account in population ethics will in the end be almost fully determined by its underlying axiology. So, for example, if your underlying axiology is totalist (and thus likely longtermist), then your overall theory will tell you to act as if you are a total utilitarian (and thus a longtermist), even if the overall theory itself is neither utilitarian nor even consequentialist.

16:00-16:40 - Bridget Williams, “Age and the value of averting death in public health”

Averting death, and thereby prolonging life, is one of the core goals of public health. However, should all deaths be considered equally important to prevent? Should public health decision-makers value averting the death of a 90-year-old differently to averting the death of a 30-year-old, or of a newborn? If so, what drives these different valuations?

How we answer these questions influences how we consider the relative importance of health issues, and the effectiveness of competing policy options. Yet, there is inconsistency in how mortality is counted and valued in public health practice. There is also persistent disagreement in the ethical literature. Although deprivationist approaches are commonly accepted, with McMahan’s Time-Relative Interest Account particularly popular.

However, deprivationist accounts have been described as necessarily person-affecting. This seems to be particularly true for the Time-Relative Interest Account, as the value of lost potential depends on the strength of its connection to an existing person. It is unclear that public-health decision-makers ought to take a person-affecting approach to value. In this talk (which represents a work-in-progress) I provide an argument for why public health decision-makers ought to take an impersonal approach to value, sketch an outline of a comprehensive impersonal approach to the disvalue of death, and apply this to the question of how to value averting death at different ages.

Wednesday 23 June - Day 1 Session B

15:20-16:00 - Johanna Thoma, “What’s wrong with pure risk paternalism?”

There is a popular idea in the recent decision theory literature that rationality is permissive under risk, that is, that two agents who value outcomes in exactly the same way and have the same credences may nevertheless rationally have different preferences over risky gambles. Such agents merely differ in their ‘pure attitudes to risk’. Once we accept this idea, a crucial question arises for those who regularly make risky decisions on behalf of others, such as healthcare professionals or policy-makers: To what extent ought we defer to an individual’s pure attitudes to risk when making risky decisions that affect her? Is there anything wrong with what I will call ‘pure risk paternalism’? I will briefly illustrate the practical implications of our answer to this question for current debates in behavioural welfare economics and social choice theory. I will then show how the answer to that question depends not only on our views of what the harm of paternalism is more generally, but also on the contentious issue of what kinds of attitudes pure risk attitudes are.

16:00-16:40 - Anders Sandberg & David Manheim, “Finite ethical value”

How much value can our decisions create? We claim that a stupendously large but finite limit exists.

The question, and answer, are critical for questions of infinite ethics and ethical uncertainty. We argue that based on our current understanding of physics, there exists an upper limit of decision-relevant value. This value is constrained by the speed of light, the physics of information, fundamental aspects of decision theory, and the ability of physical beings to place value on outcomes. The argument can handle lexicographic preferences and decisions incorporating probabilities. We finally suggest that the argument can apply, in a modified form, even given uncertainty about physical laws.

Wednesday 23 June - Day 1 Session C

15:20-16:00 - Florian Brandl, “Donor coordination: collective distribution of individual contributions”

We propose a novel mechanism design setting in which each agent contributes some amount of a divisible resource (such as money or time) to a common pool. The agents then collectively decide how to efficiently distribute the resources on a fixed number of public goods called projects. An important application of this setting is donor coordination, which allows philanthropists with different goals to find mutually accepted causes. In general, we find that no efficient mechanism can guarantee that each agent only needs to distribute her individual contribution on her most-preferred projects and that no efficient mechanism can incentivize agents to actually participate in the mechanism. On the other hand, for the important case of dichotomous preferences, these impossibilities disappear, and we show that the Nash product rule satisfies all of the above-mentioned properties. However, the Nash product can be strategically manipulated, and we settle a long-standing open question of Bogomolnaia, Moulin, and Stong (2005) by proving that no strategyproof and efficient mechanism can guarantee that at least one approved project of each agent receives a positive amount of the resource.

16:00-16:40 - Josh Tasoff, “Changing hearts and plates – interventions to reduce meat consumption”

I report on the results of two interventions to reduce meat consumption on a college campus.  In the first study, we conduct a randomized-controlled trial of pro-vegan animal-welfare pamphlets at a college campus. We observe the effect on meat consumption using an individual-level panel data set of approximately 200,000 meals. Our baseline regression results, spanning two academic years, indicate that the pamphlet had no statistically significant long-term aggregate effects. However, as we disaggregate by gender and time, we find small statistically significant effects within the semester of the intervention: a 2.4 percentage-point reduction in poultry and fish for men and a 1.6 percentage-point reduction in beef for women. The effects disappear after two months. Though we find significant effects on some subsamples in the short term, we can reject all but small treatment effects in the aggregate.

In the second study, we delivered a 50-minute lecture on how food choices affect climate change, along with information about the health benefits of reduced meat consumption. Participants in the treatment group reduced their purchases of meat and increased their purchases of plant-based alternatives after the intervention. The probability of purchasing a meat-based meal fell by 4.6 percentage points (p < 0.01), whereas the probability of purchasing a plant-based meal increased by 4.2 percentage points (p = 0.04). While the effects were stronger during the semester of the intervention, dietary shifts persisted and remained statistically significant through the full academic year.

Wednesday 23 June - Day 1 Session D

15:20-16:00 - Kevin Kuruc, “Monetizing the externalities of animal agriculture: insights from an inclusive welfare function”

Animal agriculture encompasses global markets with large externalities from animal suffering and greenhouse gas emissions. We formally study these social costs by embedding an animal inclusive social welfare function into a climate-economy model that includes an agricultural sector. The total external costs are found to be large under the baseline parameterization. These results are driven by animal welfare costs, which themselves are due to bringing animals into existence to live lives that are assumed to be worse than nonexistence. Surprisingly, the environmental costs play a comparatively small role under a wide range of empirical and ethical parameters. The model implies both that the current sector is much larger than its optimal level and that the benefits of reduced greenhouse gas emissions from substituting poultry for red meat, for example, can produce correspondingly larger animal welfare costs. Taken together, the results suggest that societies operating under inclusive welfare functions ought to be at least as concerned with improving the quality or reducing the quantity of farmed animal lives than the climate effects of these lives.

16:00-16:40 - Luisa Corrado, “The long-run effects of public investments in space exploration”

Public investment in space missions can mitigate the risk from technological catastrophes associated with climate change, pandemics or wars.  As major future space explorations are firmly rooted in the US government agenda, research into the long-run impact of a space mission is, however, still scarce. For this purpose, evidences are firstly presented where the impact of large investments in the aerospace sector (what we call aerospace shocks) are identified under narrative restrictions related to major US space missions. We then build and estimate a two-sector model in which large government investment in the aerospace sector is included explicitly. The model allows to recover the relationship between the aerospace investment and the technology spillover which supports persistent economic growth. To conduct inference in this case, we connect the model with the data using an estimation method which is able to detect the long-run effects of the large shocks associated with the periods of increased aerospace spending corresponding to the major US space missions.

Day 1 - Keynote address

17:00-18:00 - Robert S. Pindyck, “Climate (and Other) Catastrophes”

I will review what we know and don’t know about climate change, and explain why we don’t know certain things, and the nature of the uncertainty.  I will argue that what matters is chance of catastrophic climate outcome – an outcome with a very large negative impact on GDP, broadly defined.  That raises the question of how to assess the likelihood and possible impact of a climate catastrophe.  I will argue that “Integrated Assessment Models” are not useful for this, and instead we may have to rely on rough subjective estimates, or possibly expert elicitation.  I will present some results from an expert elicitation that I conducted, and discuss the implications.  Next, I will stress that other, non-climate catastrophes are possible, and they need to receive more attention than they currently do.  Lastly, I will discuss my ongoing work on valuing human lives in the context of potential catastrophes.

Thursday 24 June - Day 2 Session A

15:00-15:55 -  David Thorstad, “The scope of longtermism”

Axiological strong longtermism (ASL) claims roughly that in a large class of decision problems, the best option is one of a small class of options that are near-best for the long-term future. The scope question for ASL asks: how large is the class of decision problems satisfying ASL? Opponents of ASL claim that this class is empty, whereas some longtermists have claimed that this class is quite large, containing many or most of the decisions we make throughout our lives. I argue that the scope of ASL, while nonempty, may be more limited than often supposed. On the one hand, I argue that ASL is plausible in the case of present-day, cause-neutral philanthropic giving. On the other hand, I pose two challenges which restrict the scope of ASL: the area challenge that ASL often fails when we restrict our attention to specific philanthropic cause areas, and the challenge from option unawareness that ASL often fails when we incorporate option unawareness into decisionmaking. 

16:10-17:05 - Jacob Nebel, “Ethics without numbers”

This paper develops an alternative to the standard framework of social welfare functionals. In the standard framework, a social or overall betterness ordering is assigned to each profile of utility functions, where utilities are real numbers. Different possibilities for the measurability and interpersonal comparability of well-being are captured, in this framework, by invariance conditions on social welfare functionals. But these invariance conditions are highly restrictive and it is not clear whether they really follow from the underlying measurability/comparability possibilities with which they are associated.

The alternative framework developed in this paper cuts out the middleman of utilities, replacing them with the properties that utilities are supposed to represent. This allows us to define the measurability/comparability possibilities directly, without the use of any invariance condition, and to state social welfare functionals that violate the standard invariance conditions without requiring inadmissible information. This suggests that the invariance conditions cannot be justified in the standard way. But they do follow from a simple principle that can be motivated by some familiar considerations from the metaphysics of quantities. I conclude by considering the case for this principle.

17:20-18:00 - Joe Roussos, “A puzzle about unawareness for longtermists”

Longtermists have made extensive use of well-developed theories of empirical uncertainty, in particular Bayesian reasoning. But Bayesianism only allows for uncertainty to be represented in a limited way. Beliefs are represented with probabilities and learning is accomplished by conditionalization. So, all resolutions of uncertainty take place by updating pre-existing beliefs; agents must have priors for propositions to learn about them at later stages. In this way, Bayesianism leaves no room for agents to learn about genuinely new states of affairs. And yet, such new possibilities arise in some of the most important challenges facing us today, from climate change to existential risk. Longtermists should worry about unawareness: intuitively, we expect our beliefs about the distant future to be subject to unknown unknowns, and the more longtermist we get, the greater the potential for error due to such unknowns. So, we should have a lot of higher-order uncertainty about our judgments about the distant consequences of our actions. This seems to threaten longtermism.

On the other hand, recent results in the literature on unawareness suggest that rationality leaves no room for doing anything about this. In their forthcoming book, Beyond Uncertainty, Steele and Stefánsson prove a Reflection Theorem for awareness. This tells us that we can’t expect our credences to change when we undergo awareness changes. Of course, they may in fact change, but there is nothing to do about it now. So, there seems to be nothing we’re rationally required to do, in the face of this expectation of future awareness changes.

I reflect on what this means for longtermists. I suggest that it recommends caution when using Bayesian reasoning – arguing that we should retain the intuitive higher-order doubt in the face of the Bayesian self-assuredness.

Thursday 24 June - Day 2 Session B

15:00-15:55 -  Nicolas Treich, “Behavioral economics and animal welfare”

Improving animal welfare may be viewed as an important moral priority. A “big challenge” is to build on the small extent to which people seem to care about animals in order to improve significantly animal welfare in our societies. Besides, strong opposing economic and political forces limit the use of market, fiscal or legal instruments in favor of animals. In this presentation, I will present some recent work in behavioral economics that may help address this big challenge by better understanding people’s choices (e.g., dietary habits, donations or voting), and in turn by nudging them toward more animal-friendly choices.

16:10-17:05 - Benjamin Tereick,  “Option value, value drift and commitment”

I model the intertemporal decision problem of a longtermist whose actions may have irreversible consequences. An initial decision-maker (the ‘ancestors’) chooses an action, anticipating that decision-makers whose objectives may differ (the ‘descendants’) will act in future time periods. 

The misalignment of objectives creates an incentive for commitment while resolving uncertainty over time creates incentives for flexibility. To analyse this trade-off, I use a principal-agent model, allowing misalignment of future decision-makers to be stochastic, and introduce a notion of existential risk and of ‘severe misalignment’.

17:20-18:00 - Heather Browning, “Wild animal welfare”

With increasing focus on wild animal ethics and welfare, it has become common to depict wild animals as existing in a state of predominantly suffering, but so far this is largely speculative. In this paper, I challenge the primary lines of evidence used to support the claim of pervasive wild animal suffering and provide an alternative picture in which wild animals may often have more positive lives. However, this too is a speculative picture and I emphasise the need for further data collection using the methods of animal welfare science, to determine the extent and scope of welfare challenges and opportunities, and their effects on welfare.

Thursday 24 June - Day 2 Session C

15:00-15:55 - Peter J Hammond, “Fundamental utilitarianism and intergenerational equity with extinction discounting”

Ramsey famously condemned discounting “future enjoyments” as “ethically indefensible”. Suppes enunciated an equity criterion which, when social choice is utilitarian, implies giving equal weight to all individuals’ utilities. By contrast, Arrow (1999a, b) accepted, perhaps reluctantly, what he called Koopmans’ (1960) “strong argument” implying that no equitable preference ordering exists for a sufficiently unrestricted domain of infinite utility streams. Here we derive an equitable utilitarian objective for a finite population based on a version of the Vickrey–Harsanyi original position, where there is an equal probability of becoming each person. For a potentially infinite population facing an exogenous stochastic process of extinction, an equitable extinction biased original position requires equal conditional probabilities, given that the individual’s generation survives the extinction process. Such a position is well-defined if and only if survival probabilities decline fast enough for the expected total number of individuals who can ever live to be finite. Then, provided that each individual’s utility is bounded both above and below, maximizing expected “extinction discounted” total utility — as advocated, inter alia, by the Stern Review on climate change — provides a coherent and dynamically consistent equitable objective, even when the population size of each generation can be chosen.

16:10-17:05 - Paolo G Piacquadio, “Economic inequality: linear and non-linear measures”

The variety of inequality indices used in empirical research begs a question: which inequality measure(s) should be used? Answering this question has proven difficult as there exists no unified framework that allows researchers to directly compare the many alternative inequality indices. Our paper helps to close this gap by providing such a framework. We make four main contributions. First, we show that nearly all existing measures of inequality admit a functional representation that is “linear:” with respect to either individuals ranks (e.g., the Atkinson family or Generalized Entropy measures) or (cumulative) incomes (e.g., the generalized Gini measures). Second, we show that all these inequality measures can be interpreted and compared through the marginal rate of substitution between two types of compound progressive transfers. A “redistribution” is equivalent to taxing individuals proportionally and redistributing the revenues lump sum. Its dual, a “reassignment,” relocates individuals across the income distribution (rather than income). Linear measures of inequality impose strong restrictions on the rate of substitution between these compound transfers. Third, to relax these restrictions, we axiomatically characterize the family of non-linear measures of inequality. Among these, we single out the families of multiplicative and power inequality indices, which combine analytical tractability with rich sensitivity to the progressive transfer schemes. Fourth, we reexamine the consistency between measures of inequality and social welfare functions.

17:20-18:00 - Lisa Kramer, “A tough nudge to crack: An experimental study on moral behavioural cues to reduce animal-based food consumption”

The proportions of food that is derived from animal versus plant sources has important implications for the worldwide economy, climate change, and human health, in addition to the well-being of the animals themselves. As such, it is important to understand attitudes toward the consumption of animal products, and what might affect their evolution.  In a series of experiments, we explored a variety of interventions rooted in both the nudge literature and the moral foundations literature, to test whether we could affect participants’ beliefs and actions around the consumption of animal products. Our results showed limited success for the interventions to significantly affect beliefs, despite the wide variety of treatment conditions and the reliance on multiple, large samples. Messaging that appealed to moral values had a more sizable impact in some interventions, particularly for participants who exhibited a high degree of morality identity. Overall, our evidence suggests that attitudes toward meat consumption may be more strongly ingrained than other attitudes and behaviors to which nudges are often successfully applied, including weight loss and healthy eating more generally.

Thursday 24 June - Day 2 Session D

15:00-15:55 - Matthias Doepke, “A Soul’s View of the Optimal Population Problem”

A long-standing challenge for welfare economics is to develop welfare criteria that can be applied to allocations with different population levels. Such a criterion is essential to resolve the optimal population problem, i.e., the tradeoff between population size and the welfare of each person alive. A welfare criterion that speaks to this issue inherently requires evaluating the welfare of nonexistent people, because some people exist only in some allocations but not in others. To make progress, we consider the population problem in an environment where population is variable, but there is a fixed supply of souls, who may experience multiple incarnations over time. Rather than pondering the value of nonexistence, from the souls' perspective comparing larger or smaller populations merely involves valuing shorter or longer waits until the next incarnation. We argue that such comparisons are possible on the basis of introspection and lead to intuitive welfare criteria with attractive properties. We emphasize that one does not have to believe in reincarnation to accept the resulting criteria; rather, reincarnation serves as a metaphor to facilitate the necessary utility comparisons.

16:10-17:05 - Maya Eden, “The Cross-Sectional Implications of the Social Discount Rate”

This paper establishes a relationship between the social discount rate and the marginal social welfare weights of people from different age groups. Under standard assumptions, the social discount rate is equal to the market interest rate if and only if the distribution of consumption across age groups is optimal. Along the balanced growth path, a social discount rate that is 4.5 percentage points lower than the market interest rate implies that it is socially desirable to reduce the consumption of a 70 year-old by 1 dollar in order to increase the consumption of a 20 year-old by 10 cents.

17:20-18:00 - Drazen Prelec, “Policy and welfare by self-signaling collectives”

Diagnostic language is a staple of political argument, as shown by expressions like: “Are we (already) a society that tolerates, cares about, is doomed to X?” By bringing pre existing conditions into the picture, the rhetorical device amplifies the stakes. A decision with minor causal consequences, concerning due process, say, is presumed informative about some large X, such as whether political liberties are robust under stress. A full cost-benefit should then take into account the signal generated by the minor action, to clarify the "diagnostic risk."

I will discuss how this analysis might proceed in principle. The theoretical ideas come from a self-signaling model of individual action (Bodner and Prelec, 2002), which I briefly review. The new context is that the self-signaling agent is corporate/collective, e.g., society, humanity. With self-signaling, expressions of preference or belief cannot be taken at face value, and standard methods of decision analysis are no longer appropriate. The question then becomes how these methods might be modified or adapted.

Thursday 24 June - Day 2 Session E

15:00-15:55 - Bård Harstad, “The conservation multiplier”

Every government that controls an exhaustible resource must decide whether to exploit it or to conserve and thereby let the subsequent government decide whether to exploit or conserve. This paper develops a positive theory of this situation and shows when a small change in parameter values has a multiplier effect on exploitation. The multiplier is taken advantage of by a lobby paying for exploitation; it can also be taken advantage of by a donor compensating for conservation, but there is a fundamental asymmetry between the two. A normative analysis uncovers when compensations are optimally offered to the president, the party in power, the general public, or to the lobby group.

16:10-16:35 - Kian Mintz-Woo, “Incentives for the Long-Term(ist)”

To address long-term externalities, I propose internalizing long-term externalized costs: according to our best estimates of the long-term costs of an activity or product, this cost should be added. 

Doing so addresses motivational determinants; insofar as individuals are price-sensitive, the increased cost will lead them to be sensitive to the long-term, while not requiring intrinsic motivation. This also addresses epistemic determinants of short-termism, ignorance of the future, via a buck-passing response to epistemic short-termism. The policy-maker passes the buck to one’s preferred long-term epistemic actors, perhaps subject to some thresholds of understanding. 

The proposal is highly modular: given increased information about long-term externalities, new issues can be added as necessary. It is scale-free: through information propagating through the market, it ends up applying at any scale. And it is incentive-compatible

Implementing this policy depends on a predictable link between a measurable product or service and a given long-term threat. It thus applies to animal agriculture antibiotics and space debris. However, it may be less applicable to other longtermist cause areas like AI safety or even areas that are as yet unknown [I hope to learn more from the GPR workshop audience about this]. 

16:35-17:05 - Philipp Schönegger, “Demanding the morally demanding: gender differences in response to morally demanding charitable solicitations”

Should we confront people with morally demanding statements to perform certain actions? Specifically, should we tell people that they have a moral obligation to donate to charity? To investigate this question, we conduct an online randomized experiment via Prolific (n=2500) where participants can donate to charity. Using a between-subject design, we provide some participants with the same moral argument as to why they should donate. We then add and vary the level of the moral demandingness to donate across treatments by adding a single sentence. We find that moral arguments increase the amount donated by 51.7%. Further, we find that increasing the level of the moral demandingness does not affect the frequency or amount of donations. However, the highest level of moral demandingness (“You are morally obligated to donate”) has disparate effects on men and women. We find that 14% more women make a donation when faced with this nudge of high moral demandingness while men donate 31.9% less. These findings suggest that demanding the morally demanding may have disparate effects on different populations and may as such be applicable only in certain contexts. 

Attending the workshop

Applications to attend the 7th Oxford Workshop on Global Priorities Research have now closed.

If you are interested in presenting at future similar workshops, please email [email protected] with an outline of your proposed topic.

Schedules of previous workshops can be found on the following links: