GPI workshop on longtermism

18-19 March 2019, Oxford University

Venue

The workshop will be held at the Faculty of Philosophy at the University of Oxford, which is located in the Radcliffe Humanities Building.

Note that this event is invitation-only.

Schedule for Day 1 (Monday, 18 March)

Time	Title	Abstract
9:30	Coffee
10:00	Hilary Greaves, 'The case for longtermism'	Longtermism is, very roughly for now, the thesis that most of the value of our actions lies in the far future, so that considerations of the very long run are the primary determinant of which action is optimific, in a wide class of decision situations. If some such claim is true, it has potentially widespread practical significance. In both small-scale and large-scale decision making contexts in both the private and public spheres, the standard practice is to assume (or otherwise believe) that attempting to influence the course of the very long-run future would be intractable, so that we should instead focus on more immediate consequences. In this talk, I will attempt to articulate a plausible longtermist claim, set out as clearly as possible the case for thinking that the claim is true, and examine the extent to which the claim can be defended against various empirical, axiological and decision-theoretic assumptions.
10:45	Lennart Stern, “An extension of the Ramsey-Cass-Koopmans model with extinction risk mitigation”	We extend the Ramsey-Cass-Koopmans model by adding a risk of extinction which at each point in time is a decreasing function of extinction risk mitigation spending. We also add a state of “existential security”. Transition to this state happens with an exogenous constant probability rate and its continuation value is assumed to be exogenous. At each point in time the representative agent decides on the flow of consumption and the flow of extinction risk mitigation, the remaining output being invested. The representative agent applies pure time discounting at a constant rate and maximises his expected discounted utility. We apply the model by assuming that the world at large behaves according to the representative agent. Now we can ask: Suppose a small actor has a one-off opportunity to shift a unit of resources from consumption to investment or from consumption to extinction risk mitigation and suppose that the subsequent path is governed again by the representative agent. What impact would these interventions have in terms of total utilitarian welfare? We answer this question for the case where the world at large is at a steady state when the small actor intervenes by providing formulas and numerical calibrations.
11:00	Coffee
11:30	Julian Jamison, “Washing out”	Let us grant that the far future dwarfs the near future in terms of human welfare. Does it follow that the maximally impactful actions are ones that focus on the far future? Only if those actions can plausibly alter the future; have a roughly linear (rather than e.g. exponentially declining) probability of doing so in isolation; won’t be counter-acted (intentionally or organically) back to the status quo; and wouldn’t have happened anyway in the due course of events. I will discuss these considerations, arguing that they weaken but do not defeat the basic argument for longtermism – and perhaps more importantly that they have implications for prioritization within that general agenda.
12:00	David Thorstad, “Pure time preference: some error-theoretic remarks”	Many philosophers have a strong intuition that rationality permits holding nonzero pure time preferences. These intuitions lend default credibility to the rationality of pure time preference. But that intuitive support is defeasible by a good error theory. In this paper, I sketch some reasons to believe that many of our intuitions about the rationality of pure time preference are unreliable. I also argue that the best causal explanation for our having these intuitions casts doubt on their truth. I conclude that there is no strong intuitive support for the rationality of many forms of nonzero pure time preference.
12:30	Lunch
2:00	Lewis Ho, “Willingness to pay to avert extinctions”	How should a society determine whether or not it is worth spending money on averting extinction risks? Martin & Pindyck have show in a 2015 paper that case-by-case cost benefit analysis is not the best way to figure out whether or not to avert a catastrophe, as the willingness to pay to avert each is affected by which of the others society decides to avert. To what extent is this the case for extinction risks? I come up with a simple model to examine the ways in which the willingness to pay to avert an extinction risk changes depending on decisions regarding other extinction risks.
2:30	Philip Trammell, “Timing for Patient Altruists”	Philanthropists must decide to what extent to spend their resources on present philanthropic projects and to what extent to invest them for use on future philanthropic projects. I investigate some features of this timing problem in the case that the philanthropist is patient—i.e. has a low (including zero) rate of pure time preference. In particular, I explore the problem in some detail in the case that the philanthropist will spend her resources on direct efforts to help the global poor, and I take preliminary steps toward outlining the problem in the case that the philanthropist will spend her resources on effecting a long-term “social trajectory change”. In both cases, I find that, given standard economic assumptions, patient altruists should likely be investing most of their resources in most circumstances.
3:00	Walk
4:00	Teru Thomas, “The Asymmetry and the Long Term”	The most straightforward arguments for prioritizing extinction risk mitigation are premissed on the view that (all else equal) one ought to bring into existence people with good lives, when the alternative is that they do not exist at all. However, many people are tempted to deny this premiss, while also accepting that we ought not to bring into existence people with bad lives. Unfortunately, there are very few well-worked-out theories that take this asymmetry between good and bad lives seriously, and so it is difficult to know what the consequences are vis a vis extinction risk and more generally. In this talk I lay out the main difficulties and explain how I think the issues most plausibly pan out.
4:45	Kevin Tobia, “Legal longtermism”	This project, in very early form, seeks to investigate longtermism from a legal perspective. A first set of questions concerns legal arguments for longtermism. That is, beyond moralarguments concerning good consequences or intergenerational justice or moral duties, are there constitutional or doctrinal avenues that require or support longtermist interventions? A second set of questions concerns the merits and specifications of some legal-policy proposals that have already been offered, putatively in line with a longtermist paradigm. For example, would a "Court of Generations" or the establishment of "intergenerational trust funds" support longtermism; and if so, what are the relevant questions in elaborating the role of these institutions?
5:00	Coffee
5:30	Matthew Rendall, “Discounting and the paradox of the indefinitely postponed splurge”	Some economists and philosophers advocate discounting future costs and benefits on the ground that otherwise, we would invest nearly all our income, but never spend it. Rather than enjoying the fruits of our investment, we would always do better to re-invest them. Drawing on Derek Parfit’s typology of self-defeating theories, I will show that this argument is a red herring. Whatever other problems undiscounted utilitarianism may raise, it is neither directly nor indirectly collectively self-defeating.
6:15	Drinks
7:30	Dinner

Schedule for Day 2 (Tuesday, 19 March)

Time	Title	Abstract
9:30	Coffee
10:00	Christian Tarsney, “The Epistemological Challenge to Longtermism”	Longtermism holds that what we ought to do is mainly determined by effects of our present choices on the very far future. A natural objection to longtermism is that these effects may be nearly impossible to predict—perhaps so close to impossible that, despite the astronomical importance of the far future, differences in expected value between our present options are mainly determined by relatively short-term considerations. This paper aims to precisify and evaluate (a version of) this epistemological objection. To that end, I develop a simple model for comparing “longtermist” and “short-termist” interventions, incorporating the idea that, as we look further into the future, the effects of any present intervention become progressively harder to predict. The model yields mixed conclusions: If we simply aim to maximize expected value, and don’t mind premising our choices on minuscule probabilities of astronomical payoffs, the case for longtermism looks robust. But some prima facie plausible empirical worldviews support probability distributions over model parameters on which the expectational superiority of longtermist interventions depends heavily on these “Pascalian” probabilities. So the case for longtermism remains strong, but depends to some extent either on plausible but non-obvious empirical claims or on a tolerance for Pascalian fanaticism.
10:45	Franz Dietrich, “Forming fully Bayesian social values: the linear-geometric aggregation rule”	In the aggregation of preferences under uncertainty, one has to determine social values of outcomes as well as social beliefs about uncertain contingencies, based on the values and beliefs of the individuals. In this aggregation problem, it is desirable that social preferences be Bayesian in the full, i.e., static and dynamic, sense: society should hold expected-utility preferences, and it should revise them through conditionalizing on any event when that event is learnt, i.e., when all individuals conditionalize on it. The dynamic rationality of social preferences has received little attention, but it matters for social decision-making from a long-term perspective. I show that the class of fully Bayesian preference aggregation rules (satisfying two mild regularity conditions) consists in all rules which aggregate individual values of outcomes linearly and aggregate individual beliefs geometrically. So, the social values of outcomes should be “(possibly weighted) utilitarian”, while the social beliefs should be a (possibly weighted) geometric average of individual beliefs.
11:15	Coffee
11:45	Speed pitches on possible future longtermist research projects
12:30	Lunch
2:00	Antony Millner, “Prediction: The long and short of it”	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.
2:45	Can Baskent, “Formal Ethics: Some Ideas”	In this talk, I will discuss what foundational game theory and formal methods can introduce to the longtermism debate and what problems they can solve. First, I will construct a model for (possibly, infinite) game histories and show how subjective preferences can be defined and even updated in this model. This will allow us to express longtermist subjective preferences and introduce a dynamic element for them. Then, I will show how belief-knowledge interaction of game theoretical agents generates a Yabloesque paradox in an absolute longtermist context where we allow infinitely many players and beliefs. This approach will be useful in testing the theoretical boundaries of longtermism. Finally, I will briefly consider how digital/Internet ethics and big data can help conceptualise certain elements of longtermism beyond effective altruism.
3:15	Coffee
3:45	Owen Cotton-Barratt, "Tradable impact (for deferring to the future)"	Markets in the for-profit sector have helped efficiency. Markets in altruistic impact might do something similar. They should be particularly attractive to long-termists, since future people are likely to have better information about the longterm value of interventions made now than people today. We'll consider advantages and disadvantages of having markets in impact. We'll look at what properties would be needed in a technology to enable such markets, and consider some candidate technologies.
4:30	Lewis Ho, “Persistent effects in growth models”	Increasing economic growth is one possible way in which the long term trajectory of civilization can be changed for the better. What do growth models in economics say about which policies and shocks have effects that persist over time? In this presentation I give a brief overview of the standard economic models of economic growth, focusing on which factors they claim affect growth outcomes permanently.
4:45	Coffee
5:15	Toby Ord, “A simple framework for comparing longterm effects”	I present some simple tools for measuring and comparing very longterm effects of actions without discounting. These could help to compare the value of reducing existential risk with the longterm effects of saving a life, or with the value of speeding up progress by a year. Even if these effects can’t be modelled precisely, we can sometimes do a form of sensitivity analysis and see what the model would have to look in order for one action to be better than another. Exploring these ideas leads to some useful heuristics for thinking about longterm effects. For example, it may be useful to factor the value of a great future into three dimensions: duration, size, and quality, then look at whether the benefits of the action under consideration scale with any or all of these.
6:00	Drinks
7:30	Informal dinner