2022-11-20 Samuel Hammond Why I'm not (exactly) an Effective Altruist

Samuel Hammond 🌐🏛 (@hamandcheese): 1) I've been meaning to explain...

Why I'm not (exactly) an Effective Altruist

So here we go.



  1. EAs love to debate normative ethics.

But where do abstract moral concepts derive their motivational power?

What resources do we draw upon to know if or when to reject a vulgar utilitarian calculus?

And if those resources always existed, what's the theory for, exactly? https://twitter.com/hamandcheese/status/1594360008023719939/photo/1

  1. As Charles Taylor points out, the pragmatic force of morality exists prior to whatever framework it’s couched in.

It's instead built up from concrete social practices.

So why not ditch the moral gerrymandering and argue for a principle directly from what grounds it? https://twitter.com/hamandcheese/status/1594362275393044482/photo/1

  1. Abstract moral frameworks are just rich expressive vocabularies. They no actual justificatory work.

Joseph Heath explains this point with reference to a Mayan language that lacks modal auxiliaries and thus the word "ought."

It's a wonderful talk:


  1. Language lets us make explicit the norms implicit in social practices.

EAs mistakenly flip this order of entailment, as if the theory underwrites the practice and not the other way around — what the pragmatist philosopher Robert Brandom calls “the formalist fallacy.” https://twitter.com/hamandcheese/status/1594369669271621632/photo/1

  1. In the extreme, theories like utilitarianism reify one narrow set of commitments (reduce suffering; weigh the consequences) out of a much broader diversity of goods, resulting in a hypertrophied moral faculty that’s often indistinguishable from having no moral faculty at all. https://twitter.com/hamandcheese/status/1594370354725769220/photo/1

  2. Utilitarianism, but really any framework that axiomatizes one supreme moral good, is liable to provide ways of excusing patently unethical behavior.


  1. Moral concepts must be applied at an appropriate level of construal.

Construal level theory: We conceptualize things differently based on spatial, temporal & interpersonal distance.

Our "far mode" is abstract and idealistic. Our "near mode" is practical and particularistic. https://twitter.com/hamandcheese/status/1594373430215852033/photo/1

We need to employ each mode at its appropriate level or risk a moral category error.

In macroeconomics, the only intelligible framework is broadly utilitarian: it's the economy’s "big picture."

Yet duty and virtue still matter at the institutional and characterological levels! https://twitter.com/hamandcheese/status/1594374829272080387/photo/1

  1. An ethical life thus requires a kind of moral gestalt:

Sometimes we need to moralize about the forest, while other times we need to moralize about the trees.

Applying the far mode to near problems (or vice versa) is a category error that leads to moral pathologies: https://twitter.com/hamandcheese/status/1594376984230232066/photo/1

  1. Whether or not you're a moral realist, there are moral claims that we can all agree are false.

Accusing a deadly hurricane of murder is nonsensical, for example, since intentional properties don’t supervene on the weather.

Similar errors occur in the political domain. https://twitter.com/hamandcheese/status/1594378469382709248/photo/1

  1. Hayek famously argued that many theories of “social justice” are atavistic, as intuitive concepts of blame don’t supervene on whole collectives.

Conversely, applying a far lens on a near modality, others mistakenly conclude that we need to move “beyond blame” altogether. https://twitter.com/hamandcheese/status/1594379064223010816/photo/1

  1. At its best, the EA movement offers correctives to these category errors, pushing public policy and private philanthropy away from virtue signaling and towards a scale-appropriate sensitivity to scope.

At its worst, EAs are Charles Dickens’ "telescopic philanthropists": https://twitter.com/hamandcheese/status/1594379639765417984/photo/1

  1. EA “longtermism” all but embraces the telescope, peering far off into the distant future while our institutions crumble in the present.

As a maxim, longtermism really only makes sense for an omnipresent social planner.

All-or-nothing coin flip bets are anything but lindy! https://twitter.com/hamandcheese/status/1594380211730075648/photo/1

  1. From a longtermist perspective, EA thinking may even be an information hazard.

Actually-existing longtermist societies are oriented around order and tradition & wary of knocking down Chesterton fences.

Longtermism is thus a civilizationalist program, not a utilitarian one. https://twitter.com/hamandcheese/status/1594381024741363712/photo/1

  1. @tylercowen endorses a version of longtermism based on a zero social discount rate.

Yet in unpacking the 2nd-order implications of a zero SDR, Cowen winds up finding religion.

That is, even if longtermism is true, ordinary people probably shouldn't believe in it as such. https://twitter.com/hamandcheese/status/1594382402700328963/photo/1

@tylercowen 17) Tyler was more forthright in earlier drafts.

The (since deleted) outline from 2003 is titled “Civilization Renewed: A Pluralistic Approach to a Free Society,” and declares that “warding off decline should be a central goal, if not the central goal, of political philosophy.” https://twitter.com/hamandcheese/status/1594382925272911873/photo/1

@tylercowen 18) Stubborn Attachments is framed as consequentialist, but I prefer this draft precisely because, as Cowen notes, it avoids “being trapped by the standard difficulties of utilitarianism.”

Per @KlingBlog's Three Languages of Politics, it's also quintessentially conservative. https://twitter.com/hamandcheese/status/1594384269463654406/photo/1

  1. A longtermist commitment to economic growth has many transparently right-wing implications.

To the extent there’s a trade-off between growth and equity, we must side firmly with growth.

Trade unions, f ex., don’t just redistribute rents within a firm, but also across time. https://twitter.com/hamandcheese/status/1594385497815937024/photo/1

  1. If anything, @tylercowen argues, we should redistribute to the rich given their higher rates of savings and investment.

This is roughly the East Asian developmental model — market reforms paired with labor repression and policies to redistribute consumption into investments. https://twitter.com/hamandcheese/status/1594386339432103937/photo/1

@tylercowen 21) Similarly, longtermism implies “comparable limits on our obligations to the elderly.”

To this day, Korea stands out for its threadbare pension system and thus high rate of elder poverty.

You may not like it, but this is what peak longtermism looks like: https://twitter.com/hamandcheese/status/1594387091110109184/photo/1

@tylercowen 22) My contribution to this debate is to argue that, contra the growth-equity trade-off, robust social insurance programs are both a condition and accelerant of sustainable economic growth, reflecting the Paretian imperative to compensate the “losers” from creative-destruction: https://twitter.com/hamandcheese/status/1594387780586868736/photo/1

@tylercowen 23) The logic of a Pareto improvement makes it easily confused with utilitarianism, but the two have quite different implications.

Utilitarianism is top-down, positing a social welfare function to be maximized, while Paretians start from the bottom-up process of exchange: https://twitter.com/hamandcheese/status/1594388919013048320/photo/1

@tylercowen 24) Whereas Cowen’s defense of human rights tries to “pull a deontological rabbit out of a consequentialist hat,” a Paretian can easily reconcile our dual attachments to economic efficiency and political liberalism as derived from the common principle of mutual advantage. https://twitter.com/hamandcheese/status/1594389357032574976/photo/1

@tylercowen 25) Paretians can resolve the apparent reductios that arise from treating spatial and temporal distance as moral illusions:

Moral obligations must be appropriately construed and institutionalized in cooperative social structures, rather than derived from some cosmic standpoint. https://twitter.com/hamandcheese/status/1594390223475339264/photo/1

@tylercowen 26) In my day job at a think tank, I’m not that far off from your typical EA.

Consequentialism is pretty handy for analyzing public policy.

My work on child allowances, for ex., is directly influenced by EA thinking on the superiority of cash transfers for alleviating poverty. https://twitter.com/hamandcheese/status/1594391875355541505/photo/1

@tylercowen 27) Yet I also follow a set of professional ethics, like not stealing my coworkers’ lunch to give to the homeless even if it'd increase utility.

EAs go wrong when they embody a far conceptual mode in daily life, stripping moral obligations of their institutional embeddedness. https://twitter.com/hamandcheese/status/1594392817450852353/photo/1

@tylercowen 28) The EA movement often looks more like virtue ethics for nerds: ethical veganism, “earning to give,” the life you (specifically YOU) can save.

Yet this is all an enormous category error, as consequentialism is about integrating over system level outcomes, not intentions. https://twitter.com/hamandcheese/status/1594393491337056256/photo/1

@tylercowen 29) EAs are surprisingly blind to the Christian genealogy of their morals, believing they arrived at their beliefs through a book rather than the inherited presuppositions of the culture they grew up in.

Cowen once put this point to Peter Singer directly: https://www.youtube.com/watch?v=q4qH8wjPBz0 https://twitter.com/hamandcheese/status/1594394990045761539/photo/1

@tylercowen 30) The term “altruism” itself was first coined in the 1850s by the French sociologist, August Comte.

Comte's positivism extolled a kind of scientific naturalism but needed an ethical system to go with it. Comte thus founded a rationalist cult called the “Religion of Humanity.” https://twitter.com/hamandcheese/status/1594395398952685569/photo/1

@tylercowen 31) The “Religion of Humanity” was a proto-EA movement that sought to rid Christianity of its superstitions while retaining its moral precepts, including asceticism, a belief in “vivre pour autrui” (living for others), and a melioristic commitment to worldly improvement. https://twitter.com/hamandcheese/status/1594395781280239618/photo/1

@tylercowen 32) Yet calling EA a religion isn’t meant as a knock. As David Foster Wallace said, “Everybody worships.”

The religious structure of the EA movement may even be the best thing going for it, ensuring its high-minded ideals are embedded within a living ethical community. https://twitter.com/hamandcheese/status/1594396179013722117/photo/1

@tylercowen 33) There’s clearly an appetite among smart young people to adhere to a system — any system — that integrates their desire for social impact. So while one might prefer that EAs all became Mormon, it could be a lot worse.

At least they’re not woke! 😬 /fin https://twitter.com/Peter_Nimitz/status/1591650996916473856

@tylercowen Hat-tip to @robinhanson @AgnesCallard @bryan_caplan @slatestarcodex @KlingBlog @GarettJones @PeterSinger @Law_Rhetoric @adamgurri @SBF_FTX @carolinecapital @SamoBurja @jandrewpotter & @nntaleb for influencing this piece, either explicitly or implicitly.