Classical Ideological Audit — Effective Altruism v1.1
Classical Ideological Audit — Effective Altruism v1.1
Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
Step 0 — Protocol Activation
Instrument. Classical Ideological Audit v3.0. Commitment numbering per CPA v1.1 §IV: C1 Substance Dualism, C2 Libertarian Free Will, C3 Ethical Intuitionism, C4 Foundationalism, C5 Correspondence Theory, C6 Moral Realism.
Two guards raised before Step 1 rather than at the summary.
Failure Mode 3, Political Verdict Substitution, is at elevated risk on this target. Effective altruism is at present the subject of a live policy dispute over artificial intelligence regulation. The run audits presupposed content only. It issues no verdict on that dispute, on regulation, on any organization, and on no claim about what risks obtain. Nothing in the findings bears on whether any particular empirical or policy conclusion held by an effective altruist is accurate.
Failure Mode 5 guard: the audit runs on presupposed content. The ideology’s asserted claims — cost-effectiveness figures, probability estimates for particular outcomes, the tractability of particular causes — are recorded and not audited.
Stated condition of the run. The prose renderer is Claude, made by Anthropic, which is a party to the dispute named above. The failure-mode notes at Step 6 carry this.
Corpus sources opened in full for this version.
- Sterling, “Two and One-Half Ethical Systems,” International Stoic Forum, 18 August 2011.
- Sterling, “Stoicism and Self-Interest: What Is Our Interest?” Stoics Yahoo Group, 28 August 2014.
- Sterling, opening post to the Egoism and Altruism thread, International Stoic Forum, 11 May 2017.
- Sterling, On What Makes a Stoic.
- Core Stoicism; Nine Excerpts.
No prior conclusion stated or implied.
Step 1 — Ideology Statement
Core presuppositions, the intersection across declared variants.
P1, impartial aggregation. What is to be promoted is welfare summed across all bearers of welfare, counted without regard to proximity, relation, nationality, species or identity. One bearer’s welfare is substitutable for another’s at equal magnitude. The demand is stated on the axis running from egoism to altruism: the agent’s own welfare enters the sum at the weight of a stranger’s, and the doctrine’s distinctive requirement is a movement along that axis away from the near and the one’s own.
P2, welfare-capacity as the criterion of moral standing. What makes a thing a moral subject is its capacity to be benefited and harmed — to undergo states of welfare. Standing scales with that capacity and is extended wherever the capacity is found.
P3, maximization as the content of obligation. The right action is the one producing the greatest aggregate welfare achievable. Obligation is a maximizing problem, and falling short of the maximum is a shortfall in the action’s rightness.
P4, expected value as the decision procedure. Under uncertainty, options are ranked by probability-weighted outcome magnitude, quantified and rendered commensurable across domains on a common scale.
P5, evidence and explicit reasoning as arbiter of moral practice. Which ends to pursue and how is settled by investigation, measurement and argument, not by untutored moral response. This is the movement’s own self-definition.
Declared variants. Global health and development effective altruism, measurement-first and cost-effectiveness ranked; longtermism, with aggregation extended across time to future and merely possible welfare-bearers; animal-welfare and suffering-focused effective altruism, pressing P2 to its widest extension; the existential-risk strand with its rationalist-adjacent decision theory.
Variants excluded from the core-defining set. First, the self-description that effective altruism is not a doctrine but only the question how one can do the most good. If that framing abandons P1, P3 and P4, it abandons the foundation from which the entire cause-prioritization structure is derived; following the precedent set for a variant that abandons a load-bearing presupposition, it is ruled a departure from the system rather than a variant reading of it, and it does not soften the core. Second, and on the same ground, any virtue-ethical reconstruction making the agent’s character rather than produced welfare the object of the doctrine, which abandons P3.
Step 2 — Stage One Core Audit
Domain engagement, settled before the findings. Sterling’s 2011 division of the field governs here: consequentialism and deontology are rival answers to the question what one should do now, while virtue theory adds the requirement of settled character to one of those answers rather than supplying an independent account of correct action. Effective altruism is an answer to the action-selection question, and so is the corpus’s deontological intuitionism. The two are therefore competing at the same level, which is why no finding below is available as Orthogonal.
C1 Substance Dualism — Divergent, load-bearing
Domain engaged: P2 is a direct claim about what a moral subject is. The corpus identifies the agent with a rational faculty distinct in kind from what happens to it. P2 identifies the moral subject with a capacity to undergo states — a thing constituted by what happens to it. The contradiction is not that effective altruism asserts materialism; materialism is not in the intersection and is not attributed here. It is in the criterion: standing is conferred by sentience rather than by possession of a rational faculty, and the faculty carries no standing of its own on this account.
Load-bearing: remove welfare-capacity as the criterion and P1 loses its subject-count, the extension to non-rational animals loses its warrant, and the aggregation cannot be formed. This is the weaker of the two dissolution-relevant findings, resting on the substituted criterion of subjecthood rather than on an asserted metaphysics of mind, and it is recorded as Divergent on that narrower ground.
C2 Libertarian Free Will — Partial Convergence
Domain engaged: the doctrine addresses agents and asks them to change course. Affinity: throughout every declared variant, the mode of address presupposes a faculty that can originate a change of direction on the strength of a reason — the pledge, the career change, the argument expected to move a man who was going to do otherwise. Without such a faculty the movement’s entire advocacy is idle, and the presupposition is genuine rather than decorative.
Residual divergence: the origination presupposed is instrumental only. It is valued as a means of producing outcomes; the act of assent has no standing of its own, and the agent is assessed by states of the world he does not originate. The doctrine is additionally neutral on origination proper and consistent with a compatibilism that denies it, and the decision-theoretic strand models choice as computation over a utility function. The residual prevents Convergent; the operative and indispensable presupposition of a faculty that can be moved and can commit prevents Divergent.
C3 Ethical Intuitionism — Divergent, load-bearing
Domain engaged directly and by the doctrine’s own account of its method. P5 demotes moral apprehension from a faculty delivering knowledge to a source of defeasible data, systematically corrected where it conflicts with the aggregative calculation. The signature arguments operate precisely by exhibiting untutored moral response as inconsistent, scope-insensitive and partial to proximity, and then requiring its revision; where intuition and calculation conflict the calculation governs, and willingness to accept a conclusion the response rejects is treated within the movement as a mark of seriousness.
What is demoted can now be named precisely. Sterling identifies the smooth fit for Stoic ethics as Ross-type deontological intuitionism, in which prima facie duties are directly apprehended contextual demands requiring judgment for their application rather than rule-derived outputs, and in which the Stoic kathêkon has the same structure. These are exactly what the doctrine overrides: the duty of fidelity, the duty to one’s own child, the duty of office. Sterling’s judge case is the counterexample the finding needs — the judge asks whether he serves his social nature by enforcing the law or by acting altruistically toward the convicted man and setting him free. The welfare arithmetic cannot see the duty of office at all, and can register it only as a disposition whose consequences may be weighed. The corpus’s reading of Ross, and the CPA run on Ross already in the archive, are the cross-references here.
Load-bearing: restore the authority of direct moral apprehension at the foundation and the doctrine’s characteristic conclusions become unreachable, since each is obtained by overriding it.
C4 Foundationalism — Structural Imitation
The architecture is foundationalist and unusually explicit: a single foundational proposition — welfare is what matters, suffering is what is to be prevented — from which a derived structure of cause prioritization, career selection, institutional design and resource allocation is systematically obtained, with disputes resolved by tracing back to the foundation. The formal shape replicates C4 exactly: one foundation, a dependent structure derived from it, no appeal to vague sentiment, stopping points named.
The content at the foundational node is substituted: a welfare state occupies the position where C4 requires rational moral intuition. The imitation is architectural and the content at the point of resemblance is incompatible. This matches the pattern recorded at C4 in the Epicureanism run, with the aggregative apparatus in place of the hedonic criterion of the individual life.
C5 Correspondence Theory of Truth — Partial Convergence
Affinity, and the strongest affinity returned in any run of this instrument to date: the doctrine holds, explicitly and operatively, that beliefs are answerable to mind-independent fact and are to be revised when they fail to match it — calibration, forecasting, publication of error, refusal to let a claim stand on its appeal. This is a real commitment in the domain and not a formal resemblance.
Residual divergence: in the moral domain the propositions are handled as credences within a maximizing calculation rather than as claims true or false by correspondence to a moral fact. The moral-uncertainty apparatus converts moral propositions into a probability distribution over frameworks, and the calculation can settle an action without any moral proposition having been asserted as true. Under the corpus’s C5, which is stated for objective moral facts specifically, that residual blocks Convergent; the genuine and operative commitment to correspondence in the empirical domain blocks Divergent.
A second and independent blocking reason is supplied by Sterling directly. He objects to the equation of morality with altruism on the ground that whether an action was altruistic can, with proper finagling of definitions, be observed, while whether the altruistic action is truly virtuous or appropriate cannot. The doctrine’s evidential method reaches the observable proxy and is constitutionally unable to reach the property that settles the moral question. That is a correspondence failure within the moral domain and not a metaethical hesitation, and it does not depend on how the corpus’s C5 is scoped.
Stated exposure, reduced from v1.0. The finding no longer rests on a single reading of a single commitment. Were the corpus’s C5 read as a general theory of truth rather than as applied to moral propositions, the credence objection would fall away and the observability objection would remain, holding the finding at Partial Convergence.
C6 Moral Realism — Divergent, load-bearing
Domain engaged: P1 and P3 assert what is genuinely valuable. The corpus holds that only virtue is genuinely good, that only vice is genuinely evil, and that externals carry no genuine moral weight of their own. Effective altruism identifies the good with welfare states — precisely an external on the corpus’s account — and makes their promotion the content of obligation. The contradiction is on the identity of the good, which is the substance of C6 rather than a peripheral feature of it, and it is maximally load-bearing: if only virtue is genuinely good, and virtue is the condition of the agent’s own faculty and not transferable, then there is nothing for P1 to aggregate and the doctrine has no subject matter.
The corpus states the verdict on the target’s genus in its own words rather than by inference from the theorems. Sterling, in his 2011 division of the field, holds that insofar as consequentialism identifies the good with the external consequences to be maximized, it is fundamentally mistaken. The present finding is an application of that sentence to a doctrine whose foundation is a sum of welfare states.
The instrumental test. Sterling’s 2014 argument supplies the general principle: any conception of life that permits the goal to be identified without reference to virtue makes virtue at best a means to that goal, so that virtue is neither necessary nor sufficient for happiness. Aggregate welfare is such a goal. Within the doctrine, beneficence and the traits that produce it are instrumental, and the agent’s own condition is not what is at stake. Of Sterling’s three test cases only the first transfers, and this is stated rather than glossed: Smith, with his comfort, his freedom from care and his philanthropy undertaken to attract women, is better off on any welfare accounting, and the doctrine is committed to crediting his donations at face value, since it holds expressly that the donor’s motive does not alter the benefit done. The Ring of Gyges and the dying molester are built against egoistic instrumentalism and do not reach a doctrine whose impartiality answers them by the victims’ welfare.
The axis. Sterling’s 2017 post holds that the distinction between selfish and altruistic behaviour is itself undermined on the Stoic view, since there can be no conflict between what is good for me and what is good for you: the Stoic, like the selfish man, always seeks his own good, and, like the altruistic man, will often give up money, possessions or life itself to do what is right, without ever facing a choice between competing values. P1 states the doctrine’s distinctive demand as a position on that axis. The divergence is therefore not only that the good is misidentified but that the dimension along which the doctrine measures its own demandingness is one the corpus rules out.
A formal affinity is present and does not soften the finding. The doctrine is agent-neutral, holding its standard for all agents alike rather than indexing it to the agent’s perspective, which is a formal feature moral realism also has. That affinity is not admitted as Partial Convergence: the intersection does not require mind-independence, several strands being explicitly anti-realist, so what is retained is impartiality within a chosen framework and not objectivity. Nor is it recorded as Structural Imitation: the resemblance is a single formal property, not a replicated architecture with substituted content at its nodes.
Step 3 — Dissolution Finding
C1 Divergent, C2 Partial Convergence. Partial Dissolution.
The agent is not dissolved. What is dissolved is his standing as the subject the corpus identifies. He survives as a welfare-bearer among substitutable welfare-bearers and as a causal instrument capable of being moved by argument, but not as a rational faculty whose condition is the whole of his good. The faculty that can originate a change of course is retained and required; it is retained as a mechanism of outcome-production rather than as the seat of value. This is a materially different result from the Full Dissolution returned for Marxism, where the individual’s judgments are themselves products of a structure prior to him.
Step 4 — Structural Imitation Characterization
As stated at C4: the derivation of cause prioritization from a single value axiom is a foundationalist system in form, with a welfare state substituted for the rational-moral-intuitive foundation. The imitation is closer than Epicureanism’s, because the doctrine performs its derivations explicitly and publicly and treats the tracing of a conclusion back to the foundation as the test of the conclusion. C4 is the commitment the target most resembles in form and one of the points at which its content is furthest from the corpus.
Step 5 — Stage Two Variant Differential
Global health and development. Measurement-first practice strengthens the C4 architecture and strengthens the C3 divergence, since here the overriding of untutored response by a metric is the operating procedure. Confirm.
Longtermism. Aggregation is extended across time and to merely possible welfare-bearers. The expected-value handling of speculative outcomes strains the C5 affinity without removing it, and the substrate-independence and digital-minds line makes the welfare-subject explicitly functional, intensifying the C1 divergence. Confirm on all six.
Animal-welfare and suffering-focused. P2 at its widest extension, with welfare-capacity made the sole criterion and rationality expressly declined as a criterion. Confirm, and this is where C1’s divergence is most nearly explicit doctrine rather than an entailment.
Existential-risk and rationalist-adjacent strand. The modeling of the agent as a utility-maximizer under uncertainty places the greatest strain on C2’s Partial Convergence. The affinity nonetheless survives, because the vocabulary of persuasion, voluntary pledge and course-change on the strength of an argument is retained throughout this strand as much as any other. Confirm. Recorded conditionally: were a variant admitted that modeled assent as fully determined computation and abandoned the address to a faculty that can commit, C2 would shift to Divergent and the dissolution finding would move from Partial to Full. No declared variant does so.
Result: confirm across all six commitment findings.
Step 6 — Summary Finding
Part A — Commitment Pattern. Three Divergent (C1, C3, C6), one Structural Imitation (C4), two Partial Convergence (C2, C5). No Convergent findings. No Orthogonal findings: the doctrine engages every commitment’s domain directly, and nothing is absent by positive showing, so the absent value is unavailable under Failure Mode 4.
The deepest divergence is C6. The identification of the good with welfare is the foundation on which the C1 criterion and the C3 demotion of moral apprehension both rest, since an aggregable good requires substitutable bearers and requires that the untutored response to proximity be corrected. The nearest convergence is C5, where the commitment to belief answerable to mind-independent fact is real and operative.
Failure Mode 6 note. The distribution — no Convergent finding, two Partial Convergences, a dissolution finding weaker than Marxism’s — cannot be distinguished by the engine from a distribution arranged to look measured, and the renderer is a party to a dispute in which the target is the other side. Two specific exposures are therefore named so that a reader may test them independently. First, C5 is blocked from Convergent by two reasons which are independent of one another, so that overruling either leaves the finding standing; a reader who would move it to Convergent must defeat both. Second, C2’s Partial Convergence rests entirely on the doctrine’s mode of address to agents, and if that address is judged rhetorical rather than presuppositional, C2 goes Divergent and the dissolution finding goes Full. Both load-bearing points are on the record rather than in the verdict alone.
Part B — Dissolution Finding. Partial Dissolution, on C1 Divergent and C2 Partial Convergence. C3 and C6 Divergent independently deepen the result beyond what the synthetic rule computes, but do not convert it: the rule reads C1 and C2 only, and C2’s affinity is genuine. No declared variant shifts the finding; Stage Two confirms it throughout.
Part C — Agent-Level Implication. An individual who adopts effective altruism accepts a description of himself along with a set of claims about how to act.
He accepts that what makes him a moral subject is his capacity to undergo states of welfare, so that his standing is of the same kind as that of any other welfare-bearer and differs only in magnitude (C1). He accepts that his own direct apprehension of what is required of him is evidence to be corrected rather than knowledge to be trusted, and that when it conflicts with the calculation the calculation governs (C3). He accepts that what is genuinely valuable lies in states of the world rather than in the condition of his own faculty, and therefore that the whole of what matters is outside his control and that his moral position at any moment is a shortfall against a maximum he cannot reach (C6).
He retains a faculty that can be moved by an argument and can commit itself, and he retains the conviction that beliefs must answer to how things are (C2, C5) — which is why the doctrine is not, on this audit, a full dissolution of the agent. What is unavailable within its architecture is the corpus’s account of where his good lies: the faculty he retains is retained as an instrument of production, and no place remains for its own condition to be the thing at stake.
The two collapses. Sterling sets out the ordinary division of conduct into acts that are wrong, acts that discharge one’s duty, and acts beyond the call of duty which are admirable but optional, and holds that the third category disappears on the Stoic view, because there are no goods whose pursuit demands so much sacrifice that a good man is excused from them. Effective altruism abolishes the same category from the opposite side: its signature argument moves the saintly act out of the optional and into the required. The structural result is identical and the ground is opposite. The corpus withdraws supererogation because nothing genuinely good was ever at risk in the sacrifice; the doctrine withdraws it because something genuinely good is at risk in every dollar. An agent who has met both and noticed only that each denies him the optional has not yet found the difference that decides his condition — and the difference is the whole of it, since one of the two leaves him at every moment already in possession of what matters and the other leaves him at every moment in arrears.
This is a finding about self-description. It is not a political verdict and not a verdict on any empirical claim. Nothing in it establishes that a man holding this doctrine is mistaken about which risks obtain, which interventions are effective, or what policy is warranted; only that he cannot hold his account of what he is alongside the corpus’s.
Revision Record
v1.0 was conducted before four Sterling primary sources were opened. v1.1 adds them. No finding is altered. Five items are regrounded: the domain-engagement question at the head of Step 2, settled by the 2011 division of the field rather than by the engine’s own judgment; C3’s residual, which now names the Ross-type prima facie duties being demoted and carries Sterling’s judge case; C5’s blocking reason, which acquires a second and independent ground in the observability objection, reducing the exposure recorded at v1.0; C6, which now states the corpus’s verdict on consequentialism in Sterling’s own terms, applies the instrumental test with one of three cases marked as transferring and two as not, and adds the collapse of the egoism-altruism axis; and Part C, which gains the two-collapses section. P1 is restated to make explicit the axis on which the doctrine’s demand is positioned.
One finding remains grounded in the engine’s characterization rather than in a Sterling source: C4, Structural Imitation. A Sterling argument directed at Bentham rather than at Epicurus would be the material for it.
Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.
0 Comments:
Post a Comment
<< Home