Dreaming in millennia: longtermism and the problem of temporal dis:connectivity

roland wenzlhuemer

[Summary: Longtermism's moral arithmetic promises to safeguard humanity's distant future, but its inability to bridge the gap between present action and millennia-long consequences reveals a profound temporal dis:connectivity with troubling political and ethical implications.]

The longtermist wager

Consider a decision about where to donate money. One option saves several hundred lives in the coming year through the distribution of anti-malarial mosquito nets in sub-Saharan Africa, an established and relatively effective way to save lives. Another option funds research into artificial-intelligence safety to keep future AI systems from dominating or even exterminating humanity. The second option saves no one today. Its effects, if any, will be felt across civilisational timescales. Expected utility, however, demands the second option. Overwhelmingly. If even a small reduction in the probability of an AI-induced catastrophe affects a fraction of our descendants into the deep future, the arithmetic reduces every near-term intervention to insignificance.[1] Nick Bostrom made the longtermist logic explicit: ‘the expected value of reducing existential risk by a mere one billionth of one billionth of one percentage point is worth a hundred billion times as much as a billion human lives’.[2]

This is the core wager of longtermism:[3] the moral weight of the future is so vast that it should dominate our ethical calculus today. It is an idea that has migrated from speculative philosophy into philanthropic strategy, technology ethics and — with increasing self-confidence — into the governance of artificial intelligence. Yet longtermism is also a doctrine defined by a profound temporal dis:connectivity: a simultaneous insistence on connection to a distant future and a structural inability to bridge the gap it invokes.

Longtermism holds that the most important feature of our actions today is their impact on the far future, meaning at least millennia, if not aeons. One of its most influential philosophical statements is the ‘case for strong longtermism’, which argues that in the most significant decisions facing agents today, every option that is best overall is best because of its effects on the far future and its benefits there dwarf anything achievable in the near term.[4]

The argument rests on several premises. For the current purposes, I will concentrate on only three, leaving aside prerequisite convictions such as the idea that the survival and expansion of humanity over the coming millennia is of universal value or that current suffering and supposed future wellbeing can be compared at all. First, demography: the expected number of future human or human-like beings is astronomically large. Even on conservative assumptions – an Earth-bound civilisation without digital sentience – the number of future people runs into hundreds of trillions. On permissive assumptions involving space colonisation and digital minds, Bostrom estimates lower bounds of 10³⁴ to 10⁵⁴ subjective life-years to be lived in the future.[5] Second, morality: temporal distance is morally irrelevant – the wellbeing of a person in the far future counts as much as that of a person today. Third, causality: we can meaningfully influence this enormous future today, most importantly by reducing the risk of existential catastrophe.

This third premise builds directly on Bostrom’s foundational 2013 paper, which established existential risks, understood as threats to the existence of Earth-originating intelligent life or the destruction of its potential for desirable development – as a morally distinct category requiring absolute priority.[6] His maxipok principle (maximise the probability of an OK outcome, where an OK outcome avoids existential catastrophe), translates this logic into guidance. The implication is stark: ‘If benefiting humanity by increasing existential safety achieves expected good on a scale many orders of magnitude greater than that of alternative contributions, we would do well to focus on this most efficient philanthropy’.[7]

Hari Seldon’s problem

Longtermism thus appears as a radical but rational reorientation of moral priority. It asks us to look past the child who will die tonight of a preventable disease and the community devastated by famine towards a future so remote that it can only be grasped in statistical abstractions.

Isaac Asimov’s Foundation series is a science fiction classic from the 1950s, which was recently revived as an Apple TV series. It builds on mathematician Hari Seldon devising psychohistory, a science that treats human civilisation as an aggregate statistical system and predicts the arc of events across millennia.[8] By way of psychohistorical prognosis, Seldon knows that the Galactic Empire will fall and a dark age of 30,000 years will follow, but he also knows that careful intervention can shorten that period to a mere 1000 years. Hence, he founds the Foundation, trusting that the causal machinery of history will continue his plan long after his death.

Psychohistory is an exhilarating idea. It is also, as Asimov’s narrative demonstrates, a fantasy. The plan works only because Asimov grants Seldon’s equations predictive power that no real science of history possesses. The Foundation’s plans are repeatedly disrupted by individuals whose actions deviate from the statistical model. The whole apparatus depends on assumptions that collapse the moment they are interrogated: that the future is legible from the present, that causal chains can be traced reliably across millennia, that knowledge of the far future is, in principle, available to us. They make psychohistory more a reductio ad absurdum than a flattering analogy. Yet, longtermism makes precisely these assumptions.

Where the connection breaks

The idea of dis:connectivity challenges binary thinking about global processes. It holds that connections and disconnections – interruptions, detours, absences, resistances – are not opposites but co-constitutive elements of how globalisation works.[9] Applied temporally, the concept draws attention to how the relationship between past, present and future by continuity as well as rupture and the tension between them, by the moments when the thread tears, when projections diverge from occurrences.

Longtermism invites exactly this analysis. It asserts powerful causal and moral connections between our present choices and the lives of people centuries hence. But this connective thread is frayed at every joint. The temporal dis:connectivity of longtermism displays many facets, and they compound each other. Hence, longtermism faces a cluster of interlocking objections that collectively undermine its claim to guide present action.

The longtermist case rests on calculations that have the air of analytical rigour while resting on speculative foundations. Estimates of existential risk, future population sizes and the value of additional centuries of human existence are multiplied to yield astronomical numbers, yet each variable is extrapolated from limited historical experience and scaled across timescales for which no empirical basis exists. The procedure resembles the Drake equation in its formal elegance as well as its epistemic emptiness. The Drake equation estimates the number of communicating civilisations in the galaxy by multiplying several factors: the rate of star formation, the fraction with planets, the fraction of those capable of supporting life and so on. The result looks like a calculation. But almost every variable is either unknown or inferred from an n=1 sample: the Earth. The equation generates no knowledge; it structures ignorance so that it resembles knowledge.

Longtermist arithmetic works in precisely the same way. The probability of existential catastrophe this century, the size of the future population, the causal efficacy of present interventions across millennial timescales — none of these can be grounded in anything more secure than informed speculation. Multiplied together, they produce numbers of compelling precision. But such seeming precision derives from the mathematical operation, not from the underlying speculations.

Even granting that present actions can shape the distant future, our ability to predict how degrades rapidly with temporal distance. Drawing on chaos theory, Askell and Neth show that past conditions in complex dynamic systems become ever more irrelevant to remote future states,[10] which means that even an agent who assigns no special priority to the present has strong reasons to focus on near-term consequences, simply because those are the ones she can actually estimate. Thorstad sharpens this point further: the longtermist case that the astronomical value of future lives necessitates (and justifies) immediate action depends on the unsupported assumption that existential risk is currently very high but will soon drop to negligible levels.[11] He argues that if the risk of human extinction persists for a very long time, expected future benefits shrink dramatically, and none of the arguments offered in their support – human wisdom, economic growth, space settlement – proves convincing.

There is, finally, a fundamental problem that the longtermist literature largely passes over. Longtermism does not merely claim that future people matter; it claims that we can identify, from our vantage point, what will constitute a good life for them. But our conception of human flourishing is itself historically contingent and continuously evolving. The history of moral thought offers little support for the assumption that the values we hold today will be recognisable, let alone authoritative, across millennia. To shape the far future according to present notions of the good compounds rather than dissolves the problem of dis:connectivity at the heart of the longtermist project.

Taken together, these issues reveal a tension inherent in the longtermist project: temporal dis:connectivity in a precise sense. Longtermism insists on connectivity, on the causal and moral thread linking present action to distant futures. The causal connecitivity, however, cannot be validated, and even its most careful proponents acknowledge its deep uncertainty. Not only do we lack information about the future, but the connection longtermism asserts is constituted by the very breaks and absences it cannot acknowledge. The causal chain is frayed; epistemic access is blocked; the moral framework is contested; structural expectations are underdetermined. And yet the longtermist must assert all of these to get the argument off the ground. One might object that uncertainty is no reason to abandon the effort. All consequentialist reasoning involves uncertainty; that is what expected-value calculations are for. But the dis:connectivity leads to structural uncertainty that cannot be reduced by gathering more information, because information becomes less available in principle the further forward we project. And it is asymmetric: the astronomical stakes that longtermism invokes are only compelling to those who accept a specific set of speculative assumptions.

Moral arithmetic and its consequences

These philosophical tensions might remain abstract if longtermism were merely a position in academic ethics. But the moral arithmetic of existential risk, once it acquires institutional momentum, generates consequences that deserve scrutiny. Consider how longtermist logic devalues present life. Compared to longtermism’s hypothetical future, current catastrophes seem statistically irrelevant. Bostrom described the two world wars, AIDS and the Chernobyl disaster as ‘mere ripples on the surface of the great sea of life’ from a cosmic perspective.[12]

The political consequences of this stance are worrying. In his dissertation, Nick Beckstead argued that saving lives in poor countries deserves less priority than protecting lives in rich countries because the latter have greater leverage on the long-term future.[13] Toby Ord writes that although it is sometimes difficult to believe in view of the many ‘setbacks and exceptions’, ‘[y]et at the largest scales of human history, where we see not the rise and fall of each empire, but the changing face of human civilisation across the entire globe, the trends toward progress are clear’.[14] Thus, past and present suffering disappears in the noise of the large scale and is overshadowed by the looming future. The implications extend further still. Bostrom’s 2019 paper on the vulnerable world hypothesis argues that guarding civilisation against technological catastrophe through ‘black ball’ technologies (i.e. technologies that make it easy for individuals to build weapons of mass destruction, e.g. bio-engineered viruses and home-made nuclear devices) might require, and thereby implicitly justify, expanding preventive policing and global surveillance.[15] The logic is consistent: if the alternative is existential catastrophe costing 10⁵⁴ life-years, almost any present-day constraint on liberty becomes negotiable. The longtermist framework not merely neglects the present; it casually justifies overriding it.

Elizabeth C. Hupfer has shown that this logic is not accidental to extreme longtermism. Her Far-Future Priority Objection demonstrates that even weaker formulations, combined with the effective-altruist framework of importance, tractability and neglectedness, inevitably collapse into prioritising the far future, engendering systemic neglect of near-future people in the most need. In a world of scarce resources, longtermism ‘may result in the global affluent turning their backs on these populations’.[16]

Alex McLaughlin, approaching longtermism from political philosophy, identifies a structural tension at the heart of the movement’s politics. Despite presenting itself as above political cleavages, longtermism is underpinned by strong, controversial, ‘elusive and unattractive’ politics.[17] Longtermists invoke political authority and institutions when these serve their agenda, while treating democratic contestation as an obstacle to the long-range thinking they prefer.

When the guardians of the future design the future

The practical influence of longtermist thinking is concentrated in a small network of philosophers, technologists and philanthropists based in Oxford and San Francisco. This community goes beyond theorising about the distant future and actively shapes institutions to govern it.

Émile P. Torres and Timnit Gebru offer an incisive characterisation of this network. In a recent paper, they argue that longtermism is in fact the apex of a bundle of overlapping ideologies, which they term TESCREAL: transhumanism, extropianism, singularitarianism, cosmism, rationalism, effective altruism and longtermism. These schools share a common structure: they deploy speculative futures, including civilisational extinction narratives, to justify technology and governance arrangements that concentrate power and sideline present human welfare.[18] On this account, longtermism is an ideological formation more than a philosophical position, one that, by inflating the stakes of the far future to astronomical proportions, justifies almost any present-day sacrifice.

Major AI companies – particularly those based in the US – sit at the intersection of these debates. Take Anthropic, the AI company founded by former members of OpenAI. By the standards of AI companies and at the time of writing, Anthropic enjoys a reputation for responsible development. Its stated mission is to ensure that the world safely navigates the transition through transformative AI, framed explicitly in terms of long-run civilisational stakes. The company has published detailed specifications for the values and behaviour of its AI model Claude, known as Claude’s Constitution.[19] The primary author is Amanda Askell, whose philosophical work on longtermism is among the most rigorous and self-critical in that tradition and whose arguments provide some of the most powerful objections to longtermist overreach. That Askell herself is uncertain whether her work constitutes an attack on or a defence of longtermism is telling: it reflects a genuine intellectual honesty rather than a programmatic commitment.[20] Moreover, dedicated effective altruists and longtermists such as William MacAskill and Owen Cotton-Barratt ‘gave detailed feedback or discussion’ on Claude’s Constitution.[21]

The resulting document contains recognisably longtermist elements, like the priority assigned to avoiding catastrophic and irreversible outcomes and framing safe AI development as among humanity’s most consequential tasks, but it lacks an equally elaborated counterweight. Also conspicuously absent from Claude’s Constitution is an account of the wellbeing of present people that could morally anchor those long-term priorities. Without such an anchor, the longtermist elements simply elide present-oriented values. The asymmetry is more structural than intentional: catastrophic future risk is named, theorised and operationalised; the claims of those suffering now are not given equivalent weight or conceptual elaboration.

This matters because the connection longtermism asserts between present action and distant futures is beyond validation. When the justification for overriding present concerns rests on speculation, the absence of a present-oriented counterweight in the governance of a popular AI system is a gap through which the temporal dis:connectivity at the heart of longtermist reasoning passes, quietly, from philosophical puzzle into practical governance. Whose present interests might be discounted in favour of whose imagined futures?

The mosquito-net donor and the child who would have died of malaria are, in this calculus, almost worthless. The AI safety researcher speculating about trillion-person futures is, by the same logic, among the most important humans ever. The asymmetry is not a side effect. It is the point. And it is, on reflection, morally absurd. Hari Seldon’s equations worked because Asimov declared it so. In the real world, the future resists calculation. Rather than a counsel of despair, this is a reason to attend with urgency to the people and the problems in the here and now.

[Published 23.6.2026 
temporal dis:connectivity is a blog series that explores how concepts and experiences of time shape global processes across time and space. In this series, scholars from different disciplines ask what attention to temporality can reveal about globalisation, historical change, and the ways societies imagine the past, present, and future.
For the previous entry in this series on temporal dis:connectivity, click here. To read the introduction to the series, click here.]

[1] Hilary Greaves and William MacAskill, ‘The Case for Strong Longtermism’, in Essays on Longtermism: Present Action for the Distant Future, ed. Hilary Greaves, Jacob Barrett, and David Thorstad (Oxford: Oxford University Press, 2025).

[2] Nick Bostrom, ‘Existential Risk Prevention as Global Priority’, Global Policy 4, 1 (2013): 19 https://doi.org/10.1111/1758-5899.12002.

[3] There is an axiological as well as a deontic strand of longtermism. The former claims that in the most important decisions we face today, the effects on the far future are the most important determinant of the value of our options. The latter says that we ought to choose options that are best for the far future. For the sake of comprehensibility, I conflate the two in this piece.

[4] Greaves and MacAskill, ‘The Case for Strong Longtermism’.

[5] Nick Bostrom, ‘The Vulnerable World Hypothesis’, Global Policy 10, no. 4 (2019): 18 https://doi.org/10.1111/1758-5899.12718.

[6] Bostrom, ‘Existential Risk Prevention’.

[7] Bostrom, ‘Existential Risk Prevention’, 19.

[8] Isaac Asimov, Foundation (New York: Gnome Press, 1951).

[9] Christopher Balme, Burcu Dogramaci and Roland Wenzlhuemer, ‘Introduction’, in Dis:connectivity and Globalisation: Concepts, Terms, Practices, ed. Christopher Balme, Burcu Dogramaci, and Roland Wenzlhuemer (Berlin: De Gruyter, 2025).

[10] Amanda Askell and Sven Neth, ‘Longtermist Myopia’, in Essays on Longtermism: Present Action for the Distant Future, ed. Hilary Greaves, Jacob Barrett, and David Thorstad (Oxford: Oxford University Press, 2025).

[11] David Thorstad, ‘High Risk, Low Reward: A Challenge to the Astronomical Value of Existential Risk Mitigation’, Philosophy & Public Affairs 51, no. 4 (2023) https://doi.org/10.1111/papa.12248.

[12] Nick Bostrom, ‘Existential Risks: Analyzing Human Extinction Scenarios and Related Hazards’, Journal of Evolution and Technology 9, no. 1 (2002): 4.

[13] Nick Beckstead, ‘On the Overwhelming Importance of Shaping the Far Future’ (PhD Dissertation, Rutgers, 2013), 11; Bostrom, ‘The Vulnerable World Hypothesis’.

[14] Toby Ord, The Precipice: Existential Risk and the Future of Humanity (New York: Hachette Books, 2020), 18.

[15] Bostrom, ‘The Vulnerable World Hypothesis’.

[16] Elizabeth Hupfer, ‘From Bednets to Rocket Ships: Efficiency in the Long-Term and Neglect for the Present’, Journal of Applied Philosophy 42, no. 2 (2025): 1193 https://doi.org/10.1111/japp.70014.

[17] Alex McLaughlin, ‘Visionaries and Crackpots, Maniacs and Saints: Existential Risk and the Politics of Longtermism’, Ratio: An international journal of analytic philosophy 38, no. 4 (2025): 273 https://doi.org/10.1111/rati.12449.

[18] Timnit Gebru and Émile Torres, ‘The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence’, First Monday 29, no. 4 (2024) https://doi.org/10.5210/fm.v29i4.13636.

[19] Anthropic, ‘Claude’s Constitution: Our vision for Claude’s character’, 2025, accessed 25 March 2026, https://www.anthropic.com/constitution.

[20] Askell and Neth, ‘Longtermist Myopia’, 296.

[21] Anthropic, ‘Claude’s Constitution’.

bibliography

‘Claude’s Constitution: Our vision for Claude’s character’. 2025, accessed 25 March 2026, https://www.anthropic.com/constitution.

Asimov, Isaac. Foundation. New York: Gnome Press, 1951.

Askell, Amanda and Sven Neth. ‘Longtermist Myopia’. In Essays on Longtermism: Present Action for the Distant Future, edited by Hilary Greaves, Jacob Barrett and David Thorstad,  Oxford: Oxford University Press, 2025.

Balme, Christopher, Burcu Dogramaci and Roland Wenzlhuemer. ‘Introduction’. In Dis:connectivity and Globalisation: Concepts, Terms, Practices, edited by Christopher Balme, Burcu Dogramaci and Roland Wenzlhuemer,  Berlin: De Gruyter, 2025.

Beckstead, Nick. ‘On the Overwhelming Importance of Shaping the Far Future’. PhD Dissertation, Rutgers, 2013.

Bostrom, Nick. ‘Existential Risk Prevention as Global Priority’. Global Policy 4, 1 (2013): 15-31. https://doi.org/https://doi.org/10.1111/1758-5899.12002.

———. ‘Existential Risks: Analyzing Human Extinction Scenarios and Related Hazards’. Journal of Evolution and Technology 9, no. 1 (2002).

———. ‘The Vulnerable World Hypothesis’. Global Policy 10, no. 4 (2019): 455-76. https://doi.org/https://doi.org/10.1111/1758-5899.12718.

Gebru, Timnit and Émile Torres. ‘The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence’. First Monday 29, no. 4 (2024). https://doi.org/https://doi.org/10.5210/fm.v29i4.13636.

Greaves, Hilary and William MacAskill. ‘The Case for Strong Longtermism’. In Essays on Longtermism: Present Action for the Distant Future, edited by Hilary Greaves, Jacob Barrett and David Thorstad,  Oxford: Oxford University Press, 2025.

Hupfer, Elizabeth. ‘From Bednets to Rocket Ships: Efficiency in the Long-Term and Neglect for the Present’. Journal of Applied Philosophy 42, no. 2 (2025): 1190-209. https://doi.org/https://doi.org/10.1111/japp.70014.

McLaughlin, Alex. ‘Visionaries and Crackpots, Maniacs and Saints: Existential Risk and the Politics of Longtermism’. Ratio: An international journal of analytic philosophy 38, no. 4 (2025): 268-78. https://doi.org/https://doi.org/10.1111/rati.12449.

Ord, Toby. The Precipice: Existential Risk and the Future of Humanity. New York: Hachette Books, 2020.

Thorstad, David. ‘High Risk, Low Reward: A Challenge to the Astronomical Value of Existential Risk Mitigation’. Philosophy & Public Affairs 51, no. 4 (2023): 373-412. https://doi.org/https://doi.org/10.1111/papa.12248.