What We Owe the Future

Finished August 7, 2026.

Longtermism’s basic case

The book’s minimum commitments seem to be that people in the future matter morally, that humanity now has some opportunity to shape their future, and that—if both are granted—improving the long-run future acquires substantial moral weight. This does not require denying special obligations to family, neighbours, or one’s local community. The argument is instead that the possible scale of the future gives actions affecting it potentially disproportionate importance.

A key empirical question is whether apparent improvements are contingent on particular actors and interventions or whether they would have happened anyway. The abolition of slavery is used as an example: if British abolitionism merely accelerated an inevitable process by fifty years, its long-run effect is smaller than if slavery might otherwise have persisted into the present. Evaluating impact requires a counterfactual, not merely observing that a good change occurred.

The shard-of-glass intuition

MacAskill asks us to imagine knowing with certainty that a shard of glass outside one’s house will cut a child. It seems irrelevant whether the injury would occur tomorrow, next year, or in fifty years: one would still remove the glass. The example gives intuitive force to the proposition that temporal distance alone does not erase moral importance.

It captures something real, but also idealizes away the uncertainty and complexity that make actual long-run interventions difficult. My likely disagreement is not with longtermism’s moral premise itself, but with the move from caring about the future to deliberately redesigning the present in order to improve it.

A Hayekian knowledge problem

Wanting to shape today’s institutions for a better future does not imply that the intervention will actually improve that future. Giving actors more power to pursue a supposedly better long-run outcome may hurt more than it helps because dispersed knowledge, complex systems, and unintended consequences constrain effective planning. “We can influence the future” is not equivalent to “greater intentional control improves the future.”

There is a related Lucas-critique concern: deliberate policy changes alter the behavioural and institutional relationships on which the intervention’s predictions were based. Long-run forecasts may be especially vulnerable to endogenous adaptation, as people and systems respond to the policy and invalidate naïve extrapolation from current relationships.

This leaves a moral question to revisit: is making one’s sincere best effort itself morally good when the intervention may worsen outcomes? Good intentions or responsible effort may seem praiseworthy, but a strict consequentialist assessment turns on actual or expected results. What standard of responsibility is appropriate when long-run consequences are both enormous and deeply uncertain?

Malleability and value lock-in

MacAskill compares present-day humanity to a teenager: near the beginning of its life, unusually malleable, and capable of making choices that shape its long-run character. He also compares humanity to hot or freshly blown glass, which can be reshaped while hot but becomes harder—and riskier—to change after it solidifies. I find the two metaphors largely redundant. The glass image may add fragility, but adolescence already captures malleability and path dependence.

The discussion is strongly oriented toward AI and the possibility that technological developments could make current choices unusually consequential. That emphasis is unsurprising given the book’s proximity to effective-altruist and longtermist debates about advanced intelligence. It may be especially interesting now, but it is worth remembering that this context shapes which risks and interventions receive attention.

One of the stronger parts concerns cultural or value lock-in. If future institutions or technologies make some values extremely persistent, choices made during a malleable period could matter for a very long time. But this raises the difficult question of which culture or values should be locked in.

Lock in revisability, not a final doctrine

MacAskill’s answer, as I understand it, is effectively “nothing substantive in particular.” Moral views should remain capable of evolving because there is no reason to assume that present values are final or correct. The practical proposal is therefore to preserve institutions that permit continued moral reflection, disagreement, and revision rather than locking in one comprehensive worldview.

That is sensible and relatively unobjectionable, though not especially surprising. It also fits the Hayekian concern: procedural openness may be safer than granting planners authority to impose a supposedly optimal substantive future.

My reservation is that this position can still become self-serving. It risks implying that the author and his intellectual milieu are already several decades ahead of society and understand the direction in which moral progress will move. It may smuggle in a Whiggish picture of history—the belief that the long arc bends predictably toward justice—while presenting itself as neutral openness. I am not convinced that this confidence is warranted.

The paradox of tolerance

MacAskill appropriately raises the paradox of tolerance. Institutions committed to openness must contend with intolerant doctrines that reject pluralism and demand universal compliance. His response appears to be that this is a real problem open institutions simply have to manage rather than something with a clean theoretical solution.

That is acceptable, although a proper solution would have been more impressive. Overall: not mind-blowing, but a fine and defensible treatment so far.

A much weaker ending

The final part is by far the weakest. MacAskill plugs 80,000 Hours and advertises effective altruism in a way that feels more self-promotional than argumentative. Even the name bothers me: “effective altruism” seems to assume the success it is meant to demonstrate; efficient altruism would be a more honest description of the project.

The discussion of economic growth is especially confused. MacAskill argues that growth must eventually end because the universe contains a finite number of atoms, but a few pages later acknowledges that growth is not merely physical-input accumulation and can also come through total factor productivity. The atoms argument may constrain input growth, but it does not by itself establish the end of value-producing innovation or productivity growth.

The book then closes with sweeping claims that it is “almost always wrong to lie” and that violence is not the answer under uncertainty—which, on this account, is nearly always the relevant condition. These may or may not be defensible positions, but introducing them on the final page without doing the philosophical work is remarkably weak.

Solar subsidies and causal inference

MacAskill also credits the German Greens and German solar subsidies with helping drive down the cost of solar panels, pointing to the fact that prices fell during the subsidy period. But solar-panel prices have fallen during essentially every period since the technology’s invention. Temporal coincidence is not causal evidence, and I do not know of credible evidence supporting the strong version of this claim.

That is a particularly bad look for a writer associated with rationalism and effective altruism: the argument does not meet the evidentiary standard that intellectual milieu claims to prize.

Final reaction

What bothers me most is the sense of highly intelligent people finding intricate ways to feel morally superior by rationalizing away ordinary human morality. The result feels pretentious rather than clarifying. It strongly reminds me of Camus’ The Fall: moral performance and judgment becoming a way of elevating oneself over others.