3rd Oxford Workshop on Global Priorities Research

Talk Abstracts -

Thu 12 December, Morning Session

9.45am: Will MacAskill, “Are we living at the hinge of history?”

In the final pages of On What Matters, Volume II, Derek Parfit comments that: “We live during the hinge of history... If we act wisely in the next few centuries, humanity will survive its most dangerous and decisive period... What now matters most is that we avoid ending human history.” This passage echoes Parfit’s discussion, at the end of Reasons and Persons, of the possibility that nuclear war could cause human extinction.

But is the claim that we live at the hinge of history true? I argue that it is not. After clarifying some arguments in its favour, I give the case that the hinge of history claim is a priori extremely unlikely to be true, and that the evidence in its favour is not strong enough to overcome this a priori unlikelihood. I then discuss the implications of my argument with respect to what matters most, if we take a long-term perspective on human history.

10.30am: Kine Josefine Aurland-Bredersen, Adaptation and mitigation: Catastrophic climate change and other disasters”

This paper develops a simple economic growth framework where society is facing multiple different catastrophic threats including a climate catastrophe. Society can cope with catastrophes through adaptation, mitigation, or by adjusting the savings rate. The paper introduces two possible links between a climate catastrophe and other disasters. First, the damage of a set of other disasters may depend positively on the level of emissions. Second, investing in research and development may avert the climate catastrophe, and partially mitigate the risk of other disasters. The purpose of this paper is to investigate how different interdependencies affect the optimal policy combination, consumption and growth.

11.20am: Andreas Mogensen, “What does the future ask of us? The value of the long-term and the problem of moral demandingness”

I argue that moral philosophers have to a large extent misunderstood the problem of moral demandingness and that much of the discussion of the demandingness objection to utilitarianism has been based on false presuppositions. The key error has been to assume that utilitarianism directs us to maximize welfare within a generation by transferring resources to people currently living in extreme poverty. In fact, utilitarianism seems to rule out any obligation to help people who are currently badly off in favour of actions directly targeted at improving the long-term trajectory of Earth-originating civilization. Taking on-board this assumption forces us to view key aspects of the problem of moral demandingness in a new light.

12.05pm: Jonas Kurle, “Modelling Structural Breaks to Improve Forecasting”

Structural breaks in random variables represent a major reason for systematic forecast failure. After explaining indicator saturation methods suited to modelling structural breaks, I will discuss potential applications to global priorities research.

12.20pm: Christian Tarsney, “Non-Additive Axiologies in Large Worlds”

Is the value of a world just the sum of values contributed by each value-bearing entity in that world? Additively separable axiologies (like total utilitarianism, prioritarianism, and critical level views) say “yes”, but non-separable axiologies (like average utilitarianism, rank-discounted utilitarianism, and variable value views) say “no”. The distinction between additive and non-additive axiologies is practically important: The former support “arguments from astronomical scale” which suggest (among other things) that the far future is vastly more important than the near future, while the latter apparently do not. We show, however, that when there is a large enough “background population” unaffected by our choices, a wide range of non-additive axiologies converge in their practical implications with some additive axiology -- for instance, in the limit, average utilitarianism converges with critical-level utilitarianism and various egalitarian theories converge with prioritiarianism. We further argue that we are in fact in the relevant limit -- what is true in the limit is true of us. This means, among other things, that arguments from astronomical scale can succeed even if we reject axiological separability.

Thu 12 December, Afternoon Session

1.35pm: Daniel Heyen, “Prediction - the long and short of it” (45 minutes)

Commentators often lament forecasters’ inability to provide precise predictions of the long-run behaviour of complex economic and physical systems. Yet their concerns often conflate the presence of substantial long-run uncertainty with the need for long-run predictability; short-run predictions can partially substitute for long-run predictions if decision-makers can adjust their activities over time. So what is the relative importance of short- and long-run predictability? We study this question in a model of rational dynamic adjustment to a changing environment. Even if adjustment costs, discount factors, and long-run uncertainty are large, short-run predictability can be much more important than long-run predictability.

2.20pm: Gustav Alexandrie, “The long-term value of saving lives”

Economic analysis often focuses on the direct short-term benefits of life-saving interventions. However, by changing future population sizes such interventions may also have important long-term effects. With a sufficiently low rate of pure time preference, the expected value of these long-term effects may exceed that of the direct short-term benefits. In this talk, I use a simple growth model in a preliminary attempt to explore the relative value of the long-term and short-term effects of saving lives.

2.35pm: Johan Gustafsson, “Social Contract Theory and Future Generations”

In this talk, I shall explore the implications of Social Contract Theory for the future and the implications of the future for Social Contract Theory. Two theories that are often thought to be supported by Social Contract Theory are the Difference Principle and Average Utilitarianism. Once we take the future and future generations into account, I shall argue that neither of these theories can be supported by Social Contract Theory. I shall then explore what kind of theories could be supported by Social Contract Theory and what implications these theories have for longtermism.

4.00pm: Adam Lerner, “The Worst Off”

I argue that people with unremitting severe depression often have lives not worth living. I then argue that this generates strong reason to prefer policies that reduce the size of the human population.

4.45pm: Linus Mattauch, “Lessons from designing climate change mitigation policy”

In this 10-minute talk, I will stress two points from my work on climate change mitigation policy that may generalise to other policy fields. First, it is common knowledge among economists that the most efficient instrument for decarbonisation is a price on carbon emissions. However, citizens generally do not view policy options in the way economists discuss them. Framing a carbon tax reform around tangible benefits might therefore enhance its political support. Second, competing views on individual well-being have under-researched economic implications for low-carbon transport and food policy.

5.00pm: Gustaf Arrhenius, “Institutions for future generations”

5.15pm: Dean Spears, “Arguments for short-termism” (with Mark Budolfson)

The well-being of every future person counts, and there will probably be many future people.  But here we argue that even someone who agrees on the quantitative importance of the long-term (and, more specifically, fully agrees with the normative premises of long-termism) should largely be focused on the next few hundred years, and in many cases should act similarly or identically to somebody who is exclusively focused on this short-term (a short-term that is, of course, much longer than the horizons of typical policy practice).  One reason is that policy and political spaces are coarse: although it is not logically necessary, in actual fact policy decisions to invest in 100-year futures will be linked to policy decisions to invest in 100,000-year futures.  The more important reason is that our generations will be replaced by future generations, who will further, ignore, or reverse our projects.  Because of the biology and economic demography of human capital -- which is suboptimal and unequal in our times --- one of the best short-term projects we could promote is investment in early-life health, human capital, and human development.  Children with better early-life health are more likely to reach their physical and cognitive genetic potentials.  We should invest in children, and build institutions to do so.  This investment will improve short-term well-being, and will also have the long-term co-benefit of causing near-term future decision-makers to be healthier, smarter, and more productive than they otherwise would be.  Similar arguments may apply to investing in short-term social and political institutions.  Perhaps the human development of coming generations is a sufficient statistic for what our generation can do for the long term: we can put coming generations in a position where their preferences and capabilities will lead them to invest in long-term well-being.

Fri 13 December, Morning Session

9.30am: Christian Tarsney, “Can We Predictably Improve the Far Future?”

In this introductory talk, I set out what I take to be the central question for the second day of the workshop: Are there actions or strategies available to us that have a predictable positive impact on the far future? I describe three reasons for pessimism about our ability to predictably improve the far future, and three reasons for optimism.

10.00am: Benjamin Tereick, “Issues in moving from short- to long-term forecasting”

Most social science approaches to forecasting center around the use of proper scoring rules (Savage, 1971) which are employed to (i) evaluate (ii) incentivize and (iii) provide feedback to forecasters when the forecasted event materializes in the short-term. When moving from the short to the long-term, all three of these uses face serious challenges. In this talk I discuss existing attempts to address them and outline ideas for future work.

11.15am: David Manheim, “An Inside / Outside View on The Application of Superforecasting for Predicting and Influencing the Far Future”

Recent work on “Superforecasting” has shown that some individuals have significantly better ability to forecast short- and medium-term outcomes than others, for a variety of reasons. Because of the recency of this work, it is unclear how well this skill transfers to longer-term forecasting, but there are reasons to be both optimistic, and skeptical. In this paper, I will combine inside-view perspectives on how superforecasters think about forecasting and uncertainty with a survey of the literature on what makes forecasting more or less successful. Finally, this review and reflection will inform consideration of how long-term forecasting to influence the path of the future fundamentally differs from the scenarios where forecasting is most successful, and what can be expected in applying the approach.

In addition, I will review some characteristics and views of complexity in systems. This includes what Senge calls “dynamic complexity” where “the same action has dramatically different effects in the short run and the long” and issues with reflexivity, whereas Gall notes, “the system is altered by the probe used to test it… The probe is altered also.” It is clear that not only can the future be heavily path dependent, but that forecasts can be self-fulfilling, self-defeating, or irreducibly complex. This will inform an analysis of when each of these three cases are expected to occur, and the classes of system, and classes of forecast, where each of these dynamics is most relevant.

12.00pm: Jamie Harris, “How tractable is changing the course of history?”

For social advocacy aimed at influencing the long-term future to be plausibly cost-effective, it must be possible for thoughtful actors to deliberately change the course of history. Historical evidence provides some insight into whether this is possible. Where the course of history appears to have substantially changed, researchers can assess whether this was caused by the intentional efforts of thoughtful actors, or mostly caused by long-term trends, luck, and the deliberate but hard-to-influence actions of others. Preliminary research into the histories of the industrial revolution, World War I, the French Revolution, and the environmental movement in the US suggests that modifying these trajectory shifts would have been difficult but possible. Humanity’s long-term trajectory could also plausibly be altered through incremental, cumulative changes. Historians and social scientists have found evidence that thoughtful actors have successfully affected government policies and public attitudes in a number of ways. The limitations of this evidence will be discussed, as will opportunities for further research.

Fri 13 December, Afternoon Session

1.15pm: Kristoffer Ahlstrom-Vij, “Can Prediction Markets Forecast the Long Term?”

Prediction markets have a good forecasting track-record in a wide range of areas. But such markets can only be used when waiting for some external event to resolve the market is an option—and in the case of long time-horizons, it simply isn’t. I argue that the solution lies in deploying a novel type of self-resolving prediction market, and that recent experimental results support the following parity claim: what we know about how traditional prediction markets work can be extrapolated onto self-resolving markets. I then put that parity claim to work in the context of long-term forecasting, by noting that (a) prediction markets tend to be driven by minorities of informed traders, and (b) informed minorities exist for both short- and long-term predictions. If such experts can be engaged on prediction markets, they will tend to generate accurate long-term forecasts. Moreover, given that prediction markets tend to be driven by informed minorities, we can afford to cast a wide net when inviting participants in the hope of engaging informed traders, without thereby inviting a risk of predictive inaccuracy—which, crucially, is not the case for alternative aggregation mechanisms such as opinion polls or deliberating groups such as committees. Finally, I reflect on whether the forecasting horizon on prediction markets can be pushed significantly beyond a decade, noting that, while it cannot be ruled out in principle, the lack of evidence makes it difficult to do anything but to speculate on the matter.

2.00pm: Steven Popper, “Short-term Actions and Long-term Objectives: Escaping the Prediction Trap”

Traditional analytical support to decision making applies the machinery of optimization to rigorous statements describing current or predicted conditions. Over the long term, or when applied to systems characterized by deep uncertainty, the advisability of this practice becomes questionable and may even result in unsuspected exposure to danger. We can reconceptualize the purpose of policy analysis and so relieve it of the need to be predictive. Several methods, both qualitative and quantitative, may be used to shift the focus from prediction to what is often the real question: what set of short-term actions will be consistent with meeting our long-term objectives across many plausible futures?

2.45pm: Eva Vivalt, “How Predictions Can Go Wrong and Directions for Future Research”

We would like to use predictions about the future to inform actions today, however, we do not know the accuracy of these predictions. I present a simple model of how predictions can go wrong. I then outline a few directions for future research.

3.45pm: Ozzie Gooen, “Predictably Predictable Futures”

The last several decades have brought a large number of contributions to statistical and judgemental forecasting endeavors. The next several decades may bring significantly more.

As things get more advanced, not only will our understanding of how to predict distant events improve, but our understanding of the accuracy of these forecasts will improve as well. This should help us to resolve questions regarding when and how much to trust these forecasts.

This talk will present an overview of what we can expect if things go well and how thinking about this can inform our current actions when optimizing for the long term future.

4.15pm: Zach Freitas-Groff, “Toward an Outside View on Humanity’s Future: Implications of the Psychology of Judgment for Long-term Predictions”

To the extent decision-making relies on long-term predictions, researchers and policymakers need to estimate distant probabilities and distributions. I review literature from the psychology of decision-making on the “inside” and “outside view” approaches to prediction.

The outside view consists of base rates and reference classes, and the inside view consists in the construction of detailed scenarios, often involving substantial intuitive judgment. I document that across a wide array of domains, people make better decisions when rely on the outside view, and I argue that despite problems, this is the best approach even for predictions over extraordinarily long time horizons. I briefly outline an application of the theory.

5.00pm: Anders Sandberg, “Everything is transitory, for sufficiently large values of “transitory””

Organisations like social institutions, states and civilizations that produce coordinated outcomes across time and space tend to dissolve over time. Indeed, all ordered systems appear to have finite lifetimes according to known physics. How much should this matter for longtermism? I will review some of the empirical findings about the stability of human institutions and physical systems and what this means for achieving long-term goals. I will also look at the problem that even the laws of physics might be transient and how this affects the timescales we should consider.