Will MacAskill, a founder of the Effective Altruism (EA) movement and author of What We Owe The Future, discusses longtermism, the contingency of moral and technological progress, institutional design, and how to positively shape the far future. The conversation centers on whether moral progress is predictable or random, how much individuals and institutions matter, and what reforms could help society better represent the interests of future generations.
The contingency of moral values and the case for a “long reflection”
MacAskill argues that modern Western moral values are highly contingent—if history had gone differently (e.g., if the Nazis had won WWII or the Industrial Revolution had happened in India), we would hold very different moral beliefs and likely consider them justified.
This should make us humble: we are probably far from the moral truth, and our current values may be mediocre compared to what’s possible.
He contrasts this with technological and economic progress, which he sees as far less contingent because strong incentives drive innovation regardless of culture or individual actors.
The “long reflection” is his proposal for a prolonged period of careful moral reasoning before any irreversible decisions lock in a particular set of values.
The goal is to let reason, empathy, and debate—rather than power or historical accident—determine which values guide humanity’s long-term future.
He rejects the idea that moral progress is a pure random walk, arguing that argument and evidence do exert a weak but real force on what people come to believe.
Is moral progress fragile or self-correcting?
MacAskill pushes back against the “teenager” analogy (that humans are recklessly dismissing tradition), arguing instead that humans are too deferential to tradition and consensus, which worked in low-change environments but is poorly suited to today’s rapid transformation.
He thinks more first-principles thinking is needed, not less, and points to EA’s early COVID-19 warnings as an example of first-principles reasoning outperforming institutional inertia.
On whether we should expect our values to be above or below average: MacAskill thinks we’re probably doing better than average, citing the abolition of slavery, feminism, liberalism, and democracy as gains that could easily have been lost.
The contingency of technology and economic progress
MacAskill distinguishes sharply between moral progress (highly contingent) and technological/economic progress (much less contingent).
Historical examples like the Egyptians’ early steam engine or the late invention of semaphore show some contingency, but he argues these are on the order of centuries or millennia, not the very long term.
Post-Industrial Revolution, the incentives for innovation are so strong that most technologies would have been developed within decades regardless of who pioneered them.
He applies this to individuals like Norman Borlaug (Green Revolution): while enormously impactful, similar agricultural advances would likely have happened soon after, so his counterfactual impact is tens of millions of lives saved, not a billion.
The people who do make very long-run contingent differences are moral and ideological leaders: Marx, Muhammad, Jesus, Confucius, and abolitionist campaigners like the Quakers.
Stagnation, population, and the drivers of progress
MacAskill’s toy model of technological progress is: population × fraction devoted to R&D × research productivity.
He is concerned about declining fertility (fewer people) and potentially declining research productivity.
On whether concentrated talent (Bell Labs, Athens, Venice) matters more than total population: he acknowledges that organizational environment can multiply individual productivity, but argues that over the grand sweep of history, sheer population size and broader cultural orientation toward inquiry matter far more.
Longtermist institutional reform
MacAskill is broadly pessimistic about institutions designed to represent future generations because future people cannot lobby for themselves, creating a co-option risk.
One proposal: a randomly selected assembly with genuine legislative power, assessed retrospectively every 30 years by a subsequent assembly, with pensions tied to those assessments—creating a backward-chaining incentive structure.
Metrics for assessment could include GDP, homelessness, technological progress, and a “war index” based on superforecaster predictions of great-power conflict.
He discusses Robin Hanson’s “futarchy” (vote on values, bet on beliefs via prediction markets) as a speculative but promising idea, though he notes practical problems with liquidity and gaming.
He worries that long-termist institutional reforms could be gamed by malicious actors, drawing parallels to how environmental impact statements have sometimes been co-opted to block environmentally beneficial projects.
His current focus is on narrow, robustly good reforms (e.g., liability insurance for dangerous biolabs) rather than grand schemes to “represent the future.”
Are corporations longtermist institutions?
No. MacAskill notes that even top-200 corporations have a half-life of only about 10 years, making them surprisingly short-lived.
Universities (Oxford, Cambridge, Bologna) and religious institutions last far longer and make decisions on much longer timescales.
He discusses the “lock-in” problem: moments of plasticity (like the founding of the US Constitution or the early Christian Church) can lock in norms for centuries, for better or worse.
The US Constitution has endured remarkably well, but a constitutional amendment enshrining slavery would have been a catastrophic lock-in.
Why now is a moment of plasticity
MacAskill argues the present era is uniquely plastic for two reasons:
The world is globally interconnected with instant communication and a diversity of moral views, making it plausible that any of several value systems could ultimately dominate.
This openness could end—through a single world government, a global monoculture, or AI-enabled permanent ideological lock-in by immortal digital rulers.
He is particularly concerned that AI could enable a single ideology to persist forever, especially if global hegemony is achieved before space settlement makes control physically impossible.
He estimates something like a 10% chance of this happening within our lifetimes.
The durability of foundations and mission drift
MacAskill discusses why long-lived foundations (Rockefeller, Carnegie) have lost impact over time:
Specifying goals too narrowly (e.g., Benjamin Franklin’s bequest for blacksmith apprentices) makes them irrelevant as conditions change.
Regression to the mean: exceptionally competent founders are replaced by statistically less exceptional successors.
He sees a tradeoff between narrow specificity (which becomes obsolete) and excessive breadth (which invites mission drift or co-optation), but generally favors broader mandates on the expectation that future people will be smarter and more capable.
How good can the future be?
On Robin Hanson’s argument that digital minds (“ems”) will be at subsistence level due to Malthusian competition: MacAskill distinguishes subsistence income from well-being—ems could be at subsistence yet experience lives thousands of times better than ours if they are in bliss.
He commissioned a survey on whether people would want to relive their lives:
Roughly 9% of Indians and 13% of Americans said they would not want to live their lives again, suggesting their lives contain more suffering than happiness.
This is striking because the Indian sample was relatively well-off and the American sample relatively poor, complicating simple GDP-well-being narratives.
He uses this and other evidence (e.g., studies on how much of their conscious lives people would skip) to argue that most lives today are probably worth living, with roughly 10% of people in rich countries regarding their lives as net negative.
Contra Tyler Cowen: can we predict what matters?
MacAskill responds to the Cowen/Collison argument that predicting long-term impact is a “mug’s game” and that we should instead focus on what has consistently done good in the past (e.g., economic growth, technological progress).
He disputes this, pointing to EA’s early pandemic warnings and the success of technological forecasting (e.g., Moore’s law holding for 70+ years, predictable trends in AI capability and biotech risk).
He acknowledges that definitive evidence for who is right may take centuries, but suggests measurable near-term forecasts (e.g., can longtermist interventions demonstrably reduce AI or biotech risk?) could provide suggestive evidence within a decade.
He notes the irony that if past impact is the guide, and if changing values has historically been the most impactful thing, then the Cowen/Collison framework might actually support longtermist-style value change rather than pure economic growth.
The problems with academia
MacAskill laments that academia fails to incentivize big-picture questions (How contingent is history? Would we recover from civilizational collapse? Is the future likely to be good?) because they fall between disciplines and seem too grand or speculative.
He contrasts this with pre-institutionalized philosophy, where grand unifying theories were the norm.
He is exploring the creation of a new university that could do both education and research far better than current institutions, though he acknowledges the difficulty of building prestige from scratch.
Career and personal notes
MacAskill remains an Oxford professor because the role allows him to develop and communicate big-picture ideas, which he sees as his highest-impact contribution, while also co-founding and advising organizations (Giving What We Can, 80,000 Hours, the Future Fund).
He briefly mentions Slime Mold Time Mold’s potato diet as a genuine nutritional curiosity (potatoes are close to a superfood) but did not try it himself.
His book What We Owe the Future was released August 16, 2022 in the US and September 1, 2022 in the UK. He directs listeners to Giving What We Can (charity pledges), 80,000 Hours (career advice), and his Twitter (@willmacaskill) for further engagement.
