{"path":"research/value-weighting-decomposition.md","content":"# Can Value Weightings Be Decomposed? Specification, Sacred Values, and Six Hard Cases\n\n**Date**: 2026-07-05\n**Status**: AI-conducted research (web-sourced, with citations); exploratory input to ontology design, not a decision\n**Related**: [self-similar-decomposition-and-claim-ontology.md](self-similar-decomposition-and-claim-ontology.md), [convergence.md](../convergence.md), [depth.md](../depth.md)\n\n---\n\n## Framing: the move under investigation\n\nA standard objection to the convergence thesis (the claim that values converge when decomposed far enough) goes like this: even after two people reach full mutual understanding — same facts, same definitions, same inferences — they can still *weight* shared values differently. One weights liberty over safety; the other, the reverse. One discounts distant strangers; the other doesn't. Weightings, the objection says, are bedrock. Argument ends where weighing begins.\n\nThe move under investigation here denies that weightings are bedrock. It says: whenever someone reports a weighting, ask *\"what exactly is being weighed against what?\"* Under that question, the weighting itself splits into constituent parts — compressed empirical beliefs, a risk posture, a scope-of-moral-circle choice, an aggregation rule, and perhaps a residue — and each part is individually arguable. A \"weight\" is a mother claim: a bundled node hiding internal seams. If the move works, the objection dissolves, because what \"survives full mutual understanding\" turns out not to be a weighting at all but a small set of much more specific claims, each of which the acid of universal reason can still touch.\n\nThis document tests that move four ways: against the philosophy of weighing (§a), against the psychology of tradeoffs (§b), against six candidate genuinely-unresolvable differences via concrete decomposition sketches (§c), and finally as a practical extraction-pipeline mechanism (§d).\n\n**Summary of findings up front.** The strong version of the move — weightings fully dissolve into non-evaluative components — is false: four of the six test cases hit an irreducibly evaluative residue. The defensible version — weightings dissolve into *structured, individually arguable claims*, of which at most one or two per dispute are irreducible evaluative posits, and even those remain contestable in a different argumentative register — is strongly supported by both the philosophical and the empirical literature. The standard objection is not refuted but *transformed*: \"weightings survive full understanding\" becomes \"typed residues survive full understanding,\" which is a far weaker claim, because the residues are no longer scalar weights, they admit further argument, and they empirically account for a minority of real-world disagreement.\n\n---\n\n## (a) The philosophy of weighing\n\n### Richardson: norms get specified, not weighed\n\nHenry Richardson's specificationism is the closest thing in the literature to a direct formulation of the move. In [\"Specifying Norms as a Way to Resolve Concrete Ethical Problems\"](http://www.horty.umiacs.io/courses/readings/richardson-1990-specifying.pdf) (Philosophy & Public Affairs, 1990), Richardson contrasts three models of bringing norms to cases: *application* (deductive subsumption), *balancing* (intuitive weighting), and *specification*. His charge against balancing is exactly the founder-side intuition: to the extent a resolution rests on intuitive weightings, \"its weightings are purely intuitive, and therefore lack discursively expressible justification.\" Specification instead qualifies our commitments \"not by a quantitative weighting or discounting but instead by qualitatively tailoring our norms to cases\" — adding where, when, why, unless clauses until \"simple inspection of the specified norms will often indicate which option should be chosen.\" A specification is rationally defensible \"so long as it enhances the mutual support among the set of norms found acceptable on reflection\" — a coherentist criterion that maps naturally onto graph-wide argumentative support.\n\nRichardson pressed the point against bioethics' dominant balancing framework in [\"Specifying, Balancing, and Interpreting Bioethical Principles\"](https://doi.org/10.1076/0360-5310(200006)25:3;1-h;ft285) (Journal of Medicine and Philosophy, 2000), arguing that Beauchamp and Childress's residual \"justified balancing... would be better served by continued specification.\" [David DeGrazia (1992)](https://doi.org/10.1093/jmp/17.5.511) independently judged specified principlism \"the most promising model of justification in bioethics.\"\n\nThe critics matter too. [Carson Strong (2000)](https://doi.org/10.1076/0360-5310(200006)25:3;1-h;ft323) argued that specified principlism \"fails to provide practical assistance\" in concrete cases, because when two rival specifications are available, *choosing between them* reintroduces the very judgment specification promised to eliminate. This relocation objection is fatal to specification as an *algorithm*. It is not fatal to specification as *argumentative restructuring* — which is all the decomposition move needs. A relocated judgment is a narrower judgment: \"should the confidentiality norm be specified with an exception for imminent harm to identifiable third parties?\" is a claim with far clearer truth-conditions than \"confidentiality outweighs beneficence.\" For a platform whose product is arguable structure rather than automated verdicts, relocation is the win condition, not the failure mode.\n\n### Chang: covering values and the parity residue\n\nRuth Chang's work supplies the move's opening question and one of its residue types. Chang argues that all comparison is relative to a *covering value* — a \"with respect to what?\" that any well-formed weighing must name ([The Possibility of Parity](https://philosophy.rutgers.edu/images/documents/publications/chang-POSSIBILITY_PARITY.pdf), Ethics 2002; [Are Hard Choices Cases of Incomparability?](http://fas-philosophy.rutgers.edu/chang/hardchoicesfinal.pdf)). A reported weighting with no articulated covering value is not yet a claim; making the covering value explicit is the first decomposition step, and it is often where the dispute turns out to live.\n\nBut Chang also identifies what remains when specification is exhausted: some alternatives are *on a par* — comparable, yet neither better, worse, nor equal ([Hard Choices](https://doi.org/10.1017/apa.2017.7), Journal of the APA, 2017). Hard choices on her analysis are not artifacts of ignorance or incommensurability; they are a fourth value relation. Her positive proposal is that within parity, agents exercise a normative power to *commit*, thereby creating new will-based reasons. For the decomposition program this is a crucial calibration: a residue can be real without being a truth-claim disagreement. Two parties who decompose all the way down to a parity node are not disagreeing about the world anymore; they have made different permissible commitments. A deliberation graph can *mark* that, which is a categorically different (and more honest) output than a winner.\n\n### Keeney and decision analysis: decomposition works; elicited weights don't survive it\n\nRalph Keeney's [Value-Focused Thinking](https://www.hup.harvard.edu/books/9780674931985) (Harvard, 1992) is three decades of engineering practice built on exactly this move: values are structured into fundamental-objectives hierarchies by iterating \"What do you mean by that?\" downward and \"Why is that important?\" upward, separating means objectives from fundamental ones. Decomposing value talk into arguable structure is not speculative; it is routine professional practice in decision analysis.\n\nWhat the same field found about *weights* is the empirical heart of the matter. Elicited attribute weights systematically fail to behave like measurements of a pre-existing quantity:\n\n- **Splitting bias**: when an objective is split into more detailed sub-attributes, its total weight *increases* — the tree structure itself changes the weights ([Weber, Eisenführ & von Winterfeldt 1988](https://ideas.repec.org/a/inm/ormnsc/v34y1988i4p431-445.html), Management Science; modeled and partially corrected by [Jacobi & Hobbs 2007](https://ideas.repec.org/a/inm/ordeca/v4y2007i4p194-210.html)).\n- **Range insensitivity**: normatively, a weight must scale with the range of outcomes on its attribute; elicited \"importance\" weights barely move when ranges change ([von Nitzsch & Weber 1993](https://ideas.repec.org/a/inm/ormnsc/v39y1993i8p937-943.html); [Fischer 1995](https://scholars.duke.edu/publication/771230)).\n- **Method dependence**: direct importance ratings, swing weights, and tradeoff weights yield systematically different numbers for the same person and problem (Fischer 1995).\n\nThe verdict from the discipline that takes weights most seriously: elicited weights do not survive decomposition — they are reconstructed every time the structure changes. What is stable enough to argue about is the more specific material underneath. This is direct, quantitative support for treating a stated weighting as a lossy compression artifact rather than a bedrock datum.\n\n### Dancy: the deep challenge — reasons may have no stable weights at all\n\nJonathan Dancy's holism of reasons is the strongest philosophical challenge, and it cuts in an unexpected direction. In [Ethics Without Principles](https://doi.org/10.1093/0199270023.001.0001) (Oxford, 2004), Dancy argues that \"a feature that is a reason in one case may be no reason at all, or even the opposite reason, in another\" — valence itself switches with context, so there are no invariant weights to decompose *into*. If Dancy is right, the four-part decomposition schema cannot terminate in context-free atomic contributions, because the parts' contributions vary holistically too.\n\nTwo things blunt the challenge. First, Dancy is an ally against the objection the move targets: he denies that stable bedrock weightings exist at all, which is more radical than the decomposition claim, not less. Second, holism is a thesis about *contributory reasons*, not about the arguability of claims. \"In circumstances C, the fact that the act damages health counts against it\" is itself a claim with truth-conditions; Dancy's own machinery of enablers, disablers, and intensifiers is a vocabulary for making context explicit — which is specification by another name. The honest cost is architectural rather than fatal: a holism-respecting graph must context-index its reason claims, which threatens combinatorial growth and weakens the transferability of accumulated general lemmas. Default-valence generalism (the view Dancy resists but many holists accept) is the pragmatic middle: treat reason-claims as defeasible defaults and let challenges carry the context.\n\n### Williams and Anderson: entangled concepts and plural value-kinds\n\nTwo further challenges constrain the *schema* rather than the move. Bernard Williams's thick ethical concepts — cruel, courageous, treacherous — are simultaneously \"action-guiding\" and \"guided by the world\" ([SEP: Thick Ethical Concepts](https://plato.stanford.edu/entries/thick-ethical-concepts/); [IEP overview](https://iep.utm.edu/thick-co/)). On the non-reductive reading, their evaluative and descriptive content cannot be factored into a neutral description plus an attitude. Where a weighting is mediated by thick concepts (\"this is *desecration*\"), a decomposition into \"empirical belief + preference\" is not lossless; the concept resists the split.\n\nElizabeth Anderson's [Value in Ethics and Economics](https://www.hup.harvard.edu/books/9780674931909) (Harvard, 1993) presses pluralism about value-*kinds*: goods differ in the *mode* of valuation they merit — use, respect, love, honor, appreciation — and some comparisons are constitutively blocked: to price a friendship is partly to fail to have one ([SEP: Incommensurable Values](https://plato.stanford.edu/entries/value-incommensurable/)). Anderson resists any single scale. But note what her view does to a weighting dispute: it converts \"how much is X worth relative to Y?\" into \"what mode of valuation does X merit, and what does that mode require here?\" — which are arguable claims about fittingness, not ineffable quanta. Anderson blocks scalar reduction while *supporting* re-expression into structured, contestable form.\n\n**Section verdict**: the decomposition-friendly tradition (Richardson, Chang's covering-value requirement, Keeney, the MCDA weight-instability findings) establishes that raw weighting-talk is under-specified, unstable, and improvable by structured re-expression. The challengers (Dancy, Williams, Anderson) do not rescue bedrock scalar weights — none of them defends the thing the objection needs. What they show is that the decomposition schema cannot be one fixed four-part formula: it must admit qualitative re-description (mode-of-valuation claims, thick-concept applications, context-indexed reason claims) as decomposition outputs alongside empirical beliefs and formal parameters.\n\n---\n\n## (b) The psychology of tradeoffs\n\n### Weights are constructed at elicitation time\n\nPaul Slovic's landmark synthesis [\"The construction of preference\"](https://doi.org/10.1037/0003-066x.50.5.364) (American Psychologist, 1995; book-length treatment [Lichtenstein & Slovic 2006](https://www.cambridge.org/core/books/construction-of-preference/994FE8DFB8D431338B2A009F25271FBC)) shows that normatively equivalent elicitation procedures produce systematically different preference orderings: preferences are \"constructed — not merely revealed — in the process of elicitation.\" [Fischhoff, Slovic & Lichtenstein (1979)](https://doi.org/10.21236/ada075100) called the underlying phenomenon *labile values*: for unfamiliar tradeoffs, \"the elicitation procedures used may be a major factor in shaping the values of people who do not hold well-formulated values.\" Together with the MCDA biases above, the empirical picture is consistent: what people stably carry around is not a weight vector but a loose set of partially articulated considerations, exemplars, and commitments. A claim-graph is a *more* psychologically realistic representation of a person's evaluative state than a utility function — the \"weight\" was never in the head in the first place.\n\nThere is also direct evidence that the \"compressed empirical belief\" component is load-bearing. [Liu & Ditto (2012)](https://doi.org/10.1177/1948550612456045) found that moral evaluations and factual beliefs cohere tightly and *causally interact*: the more inherently immoral people judged capital punishment, the less they believed it deters — and reading purely moral essays *changed factual cost-benefit beliefs* with no factual content supplied. Much of what presents as a value-weighting difference is motivated empirical disagreement wearing a values costume. Correspondingly, closing factual gaps can move evaluative attitudes: a large experiment on gun policy found that teaching cross-cutting facts produced genuine belief updating and [attitude moderation](https://pmc.ncbi.nlm.nih.gov/articles/PMC12019190/).\n\n### Protected values: the \"weight\" is often a rule, not a quantity\n\n[Baron & Spranca (1997)](https://doi.org/10.1006/obhd.1997.2690) characterized *protected values* — values people treat as blocked from tradeoffs — and found a diagnostic signature: quantity insensitivity, agent relativity, moral obligation, anger at the mere question, and \"denial of the need for trade-offs through wishful thinking.\" Their structural diagnosis is important for the decomposition program: protected values behave like *deontological rules about actions*, not like large scalar weights on outcomes. [Ritov & Baron (1999)](https://doi.org/10.1006/obhd.1999.2839) tied them to omission bias (tolerating larger harms from inaction to avoid smaller harms from action). Baron's later work adds that protected values \"are usually maintained by lack of reflection of the sort that would see counterexamples to their absoluteness\" — probing with counterexamples softens many of them ([Baron, Analyse & Kritik](https://www.analyse-und-kritik.net/Dateien/5dfbb50bf01d0_baron.pdf)). The decomposition-relevant lesson: when someone \"weights liberty infinitely,\" the correct decomposition output is often not a large number but a *re-typing* — a side-constraint claim with a scope, an act/omission asymmetry, and exception conditions, each independently arguable. (A complication for tidiness: [Tanner & Medin 2004](https://doi.org/10.3758/bf03206481) found protected-value holders were sometimes *less* susceptible to framing effects — protected values are not uniformly noise.)\n\n### Sacred values: decomposition experienced as violation\n\nThe Fiske–Tetlock research program is the decisive literature on the *felt experience* of tradeoff analysis. [Fiske & Tetlock (1997)](https://doi.org/10.1111/0162-895x.00058) argued that taboo tradeoffs transgress relational schemata: comparing a sacred value against a secular one (life vs. money) is experienced as *constitutively* degrading, because holding the value partly consists in refusing the comparison. [Tetlock et al. (2000)](https://doi.org/10.1037/0022-3514.78.5.853) documented the behavioral signature: moral outrage plus moral cleansing, and — most important for platform design — the **mere-contemplation effect**: a decision-maker who merely *deliberates longer* over a taboo tradeoff is judged more punitively, *even when they ultimately affirm the sacred value*. Yet the same deliberation over a *tragic* tradeoff (sacred vs. sacred) makes the decision-maker seem wiser. The taboo attaches to the framing, not to careful thought as such.\n\nField evidence sharpens both edges. [Ginges, Atran, Medin & Shikaki (2007, PNAS)](https://www.pnas.org/doi/abs/10.1073/pnas.0701768104), working with Palestinian and Israeli participants, found that adding *material incentives* to a proposed compromise over sacred issues **increased** violent opposition, while the adversary's *symbolic concessions* over their own sacred values decreased it (see also [Atran, Axelrod & Davis 2007, Science](https://websites.umich.edu/~axe/research/Atran%20Axelrod%20Davis%20Sacred%20Barriers%20Sci%2007.pdf)). Sacredness is real in behavior, not just rhetoric. And yet: [Tetlock (2003)](https://gwern.net/doc/philosophy/ethics/2003-tetlock.pdf) reports that people \"often acquiesce when secular violations of sacred values are rhetorically reframed as routine or tragic trade-offs,\" and [Tetlock, Mellers & Scoblic (2017)](https://www.aeaweb.org/articles?id=10.1257%2Faer.p20171110) document how much everyday sacredness is *pseudo-sacred* — abandoned under good arguments, tempting inducements, or discreet framing.\n\n**What this means for the move.** Feasibility: strongly supported — the weights are constructed, so decomposition destroys nothing that was ever stably there; and much \"sacredness\" is pseudo-sacred, i.e., a rhetorical posture over an underlying structure that *can* be engaged. Felt experience: decomposition of genuinely sacred values is itself morally coded as violation. The act of laying \"how much is dignity worth?\" on an analytic table is the taboo. But the same literature shows the escape route: what is taboo is *scalar comparison with the profane*, not structured moral reasoning as such. A decomposition that asks \"what does honoring this value require here, given these empirical beliefs?\" is specification in Richardson's sense — and lands in the tragic/routine frame that people not only tolerate but respect. One further design-relevant result: across 15 studies, [Kubin et al. (2021, PNAS)](https://par.nsf.gov/biblio/10312451-personal-experiences-bridge-moral-political-divides-better-than-facts) found opponents respect moral positions grounded in personal experiences (especially of harm) more than positions grounded in facts — some \"weights\" decompose into compressed *testimony*, and the graph should be able to hold that as first-class content rather than flattening it into propositions that strip its evidential and relational force.\n\n---\n\n## (c) The formal test: six decomposition sketches\n\n**Schema.** A reported weighting W (\"V1 outweighs V2 in context C\") is re-expressed as: (1) *empirical beliefs* about what the options lead to or consist in; (2) a *risk posture* over uncertainty; (3) a *moral-circle scope* choice (whose interests count, and when); (4) an *aggregation rule* (sum, average, maximin, thresholds, side-constraints, lexical priorities); (5) a *residue*. The test: run each of the six cases a few levels deep and name precisely where, if anywhere, decomposition stalls.\n\n### Case 1: Retribution — is deserved suffering intrinsically good?\n\nDecompose \"retribution outweighs mercy/prevention\":\n\n- **Empirical**: deterrence efficacy, incapacitation, recidivism, brutalization effects, error rates (innocents punished), costs. Note Liu & Ditto's finding above: deterrence beliefs track prior moral stance on capital punishment, so this layer carries far more of the actual public disagreement than participants realize.\n- **Risk posture**: irreversibility under error (execution vs. reversible sentences) — a risk-function question, not a values question.\n- **Scope**: does the *offender's* suffering count negatively (as everyone's suffering does), at zero, or positively? Do victims' retributive satisfactions count?\n- **Aggregation**: is punishment governed by goods-summing, or by desert as a side-constraint (both a license *and* a cap on punishment)?\n\nStrip all four and a residue remains: **the fittingness claim** — *culpable wrongdoing makes the wrongdoer's suffering non-instrumentally fitting/good*. This is exactly the claim [Michael Moore](https://doi.org/10.1017/cbo9780511625411.008) defends (\"a retributivist punishes because, and only because, the offender deserves it\"), grounding it in the epistemic credentials of retributive emotions (the first-person guilt test) and defending those emotions against the genetic fallacy. The residue is real, narrow, and precisely stateable — and it is still not conversation-terminal: it is attacked by genealogical debunking (retributive intuitions as evolved vengeance machinery — the style of argument in [Greene's \"Beyond Point-and-Shoot Morality\"](https://psychology.fas.harvard.edu/files/psych/files/beyond-point-and-shoot-morality.pdf), itself contested, e.g., [Lott 2016](https://doi.org/10.1086/687337)) and defended by fittingness vindications ([Bennett's overview](https://eprints.whiterose.ac.uk/id/eprint/95450/2/WRRO_95450.pdf)). Argument does not stop at the residue; it changes register — from weighing to genealogy-versus-vindication.\n\n**Verdict: hits a bedrock-shaped residue** — \"culpability makes suffering fitting\" — but the compression achieved is enormous: most live policy disputes (death penalty, sentencing) move substantially at layers 1–4 without ever touching it. *Confidence: high on both halves.*\n\n### Case 2: Partiality — loving your own child 100× more\n\nDecompose \"my child's welfare outweighs a stranger's child's 100:1\":\n\n- **Empirical**: children flourish under secure attachment; a practice of universal parental impartiality would predictably produce worse outcomes for children in aggregate. This grounds the classic two-level move: partiality as the impartially-optimal *decision procedure*. On this analysis a large chunk of the multiplier is an empirical claim about institutional design, fully arguable.\n- **Scope**: none needed — the stranger's child counts fully.\n- **Aggregation**: agent-relative permissions and *associative duties* versus agent-neutral value. The magnitude question (why 100× rather than 10× or 1000×?) decomposes almost entirely into role-obligation content: what specifically do parents owe, and what may they refuse to sacrifice? ([Lange's Philosophy Compass survey](https://compass.onlinelibrary.wiley.com/doi/full/10.1111/phc3.12860) maps this structure.)\n- **Deeper structure**: the ground of partial reasons is itself a mapped debate — the projects view (partiality flows from the agent's ground projects), the relationships view ([Kolodny](https://www.mit.edu/~shaslang/mprg/KolodnyLVR.pdf): the relationship is the reason; refined for parents and children in [Kolodny 2010](https://onlinelibrary.wiley.com/doi/10.1111/j.1088-4963.2009.01173.x)), and the individuals view ([Keller 2013](https://press.princeton.edu/books/hardcover/9780691154732/partiality)). Each is a structured, criticizable analysis.\n\nResidue: after the instrumental component and the role-content are stripped, one question stands — **do underived agent-relative reasons exist, or is all partiality ultimately answerable to an impartial point of view?** Notice what happened: the residue is no longer a *weight* (a multiplier) but a *structural* question about the logic of reasons. The \"100×\" was never a bedrock quantity; nobody defends the number as such.\n\n**Verdict: partially decomposable; the magnitude fully dissolves; residue = the existence of underived agent-relative reasons** (a question about normative structure, not a quantum of love). *Confidence: high.*\n\n### Case 3: Population ethics — creating happy people vs. helping existing ones\n\nDecompose \"helping an existing person outweighs creating an equally happy new one\":\n\nThis is the showcase, because academic philosophy has already run the decomposition to completion. The vague weighting unfolds into [Narveson's slogan](https://philpapers.org/rec/NARMPO) (\"we are in favour of making people happy, but neutral about making happy people\"), which unfolds into the [Procreation Asymmetry](https://en.wikipedia.org/wiki/Asymmetry_(population_ethics)) (strong reasons against creating miserable lives; no or weak reasons to create happy ones), which unfolds into crisp sub-claims, each with its own literature: *can existence be better for a person than non-existence?* (comparativism about existence); *do complaints require a complainant?*; *are reasons to confer well-being conditional on a person's existence?* ([Frick 2020](https://onlinelibrary.wiley.com/doi/10.1111/phpe.12139)); *is betterness person-affecting or impersonal?* ([Greaves's overview notes](https://users.ox.ac.uk/~mert2255/teaching/grad/TT15/population-ethics.pdf); [Grill on asymmetric axiology](https://kallegrill.se/wp-content/uploads/Asymmetric-Population-Axiology-preprint.pdf)).\n\nAnd the decomposition has a proven floor: [Arrhenius's impossibility theorems](https://www.iffs.se/media/2264/an-impossibility-theorem-for-welfarist-axiologies-in-ep-2000.pdf) show that no welfarist axiology satisfies a handful of individually compelling adequacy conditions (Dominance, Addition, Non-Extreme Priority) while avoiding the Repugnant, Sadistic, and Anti-Egalitarian Conclusions. The residue is not an opaque weight: it is a **formally mapped inconsistent polygon of precise axioms, with a theorem proving you must give one up**. The remaining disagreement — *which* axiom to surrender — is parity-like (a choice among precisely stated, individually defensible principles), and every candidate surrender is itself an argued position.\n\n**Verdict: fully decomposable into axioms; residue = axiom choice under proven joint inconsistency** — bedrock-shaped, but maximally legible. This is the existence proof for what a mature weighting-decomposition looks like. *Confidence: high.*\n\n### Case 4: Purity/sanctity — the never-discovered corpse desecration\n\nDecompose \"desecrating a corpse nobody ever learns about is wrong\":\n\n- **Empirical**: hidden harm beliefs. The moral-dumbfounding studies ([Haidt, Koller & Dias 1993](https://sdimakis.github.io/moral_psychology/readings/week_3/Optional/Haidt_1993.pdf); [Haidt & Björklund](https://polpsy.ca/wp-content/uploads/2019/05/haidt.bjorklund.pdf)) show people confabulate harms (disease, discovery, trauma) and often maintain condemnation after each is stipulated away.\n- **Scope**: do the dead hold interests (posthumous harm)? Does the community hold an interest in the integrity of its death-norms even unviolated-in-public? Both are structured, arguable positions.\n- **Character/risk**: what does willingness to desecrate do to the agent's dispositions toward the living? Norm-erosion and slippery-slope claims — empirical.\n- **Aggregation/re-typing**: an expressive claim in Anderson's mode-of-valuation sense — the act *expresses* disrespect for humanity regardless of consequences; wrongness of expression rather than of outcome.\n\nResidue candidates, and here the literature genuinely splits three ways: (i) a **fittingness claim** — some acts are intrinsically degrading independent of any experiencing subject; (ii) a **debunking verdict** — purity intuitions are a disgust program misfiring outside its evolved domain, so after decomposition *nothing* justified remains; (iii) a **hidden-harm verdict** — the dumbfounding evidence itself is weaker than advertised, and condemnation tracks implicit harm perception after all ([Guglielmo 2018, \"Unfounded dumbfounding\"](https://www.sciencedirect.com/science/article/abs/pii/S0010027717302202); cf. [Gutierrez & Giner-Sorolla 2007](https://doi.org/10.1037/1528-3542.7.4.853) on anger vs. disgust and presumed harm), in which case the residue is *empirical*. The residue node is thus itself contested territory rather than a stable atom — which, notably, is precisely what \"nothing is permanently atomic\" predicts.\n\n**Verdict: partially decomposable; residue = a three-way live dispute (fittingness posit vs. debunkable affect program vs. implicit harm belief)** — the one case where the residue's own *type* is unresolved. *Confidence: medium; the empirical literature is genuinely unsettled.* This is also the case where decomposition-as-desecration bites hardest: analyzing sanctity on a dissection table is itself coded as impure.\n\n### Case 5: Risk temperament — the certain good vs. the 10% moonshot\n\nDecompose \"a certain modest good outweighs a 10% chance of something vastly better\":\n\n- **Empirical**: is the 10% credible? Model uncertainty about the probability itself.\n- **Outcome valuation**: diminishing marginal value of the good — the orthodox expected-utility move that absorbs much apparent risk aversion into the utility curve; arguable case by case.\n- **Scope**: who bears the downside? Choosing a gamble for oneself differs from imposing it on others; a great deal of policy-level \"risk temperament\" disagreement is actually a separateness-of-persons and aggregation disagreement in disguise.\n- **Risk posture proper**: [Lara Buchak's risk-weighted expected utility theory](https://doi.org/10.1093/acprof:oso/9780199672165.001.0001) (Oxford, 2013) proves the decomposition formally — a rational agent's preferences factor into *beliefs*, *utilities*, and a third, separable component, the **risk function r(p)**, which weights how much worse-case scenarios matter relative to better-case ones. The \"temperament\" is literally a plottable mathematical object, provably distinct from both probability assignments and utility curvature.\n\nResidue: the risk function itself. Buchak argues a range of risk functions is *rationally permissible* (a permissive band, not a point), while expected-utility orthodoxy fights back with dynamic-consistency and long-run arguments. The residue is a choice within a formally characterized permissible zone, with the zone's boundaries under active, structured dispute.\n\n**Verdict: fully decomposable, with a formal residue = the choice of risk function within a contested rationally-permissible band.** The weighting has become mathematics plus a parity-like commitment. *Confidence: high.*\n\n### Case 6: The untouched — wilderness vs. the engineered paradise\n\nDecompose \"pristine wilderness has value no engineered replacement can restore\":\n\n- **Empirical**: can engineering actually replicate ecosystem function, complexity, resilience? Unknown-unknowns and option value — largely a *risk posture about irreversibility* wearing metaphysical clothes. Environmental economics already routinely decomposes \"total value\" into use, option, existence, and bequest components — arguable structure again.\n- **Scope**: do nonhuman organisms and systems hold morally considerable interests directly, or only via human valuers? A mapped, structured debate.\n- **Aggregation**: option value and existence value under irreversibility — much of the \"deep ecology vs. engineering\" heat is a risk-posture disagreement about irreversible loss under model uncertainty (layer 2 again).\n\nResidue: [Robert Elliot's \"Faking Nature\"](https://doi.org/10.1080/00201748208601955) (Inquiry, 1982; [text](https://rintintin.colorado.edu/~vancecd/phil308/Elliott.pdf)) names it precisely: \"genesis is a significant determinant of an area's value.\" A perfectly restored ecosystem stands to the original as a flawless forgery stands to the artwork — indistinguishable in intrinsic properties, yet lacking a value that depends on *causal history*. Decompose the genesis claim one more level and it splits: (i) the general claim that historical and relational properties can be value-relevant — which is *widely shared*, not exotic (the forgery case itself; achievements vs. lottery wins; the near-universal refusal of Nozick's experience machine); and (ii) the specific claim that **causal independence from human intention is one such value-conferring property**. Claim (ii) is the residue. It remains attackable — via the sorites pressure that naturalness admits of degrees and nothing is fully pristine, and via the debunking suggestion that \"untouchedness\" projects a human romance onto indifferent processes (the restoration debate around Elliot and Eric Katz's \"The Big Lie\" is mapped in [Light's review](https://shs.hal.science/halshs-00335250v2/file/8.Light.pdf)) — and defensible in turn. Note what decomposition bought: the dispute shrank from \"ineffable nature-mysticism vs. Promethean engineering\" to \"does the authenticity-value everyone already grants in the art and achievement cases extend to ecosystems?\" — a bridgeable question, because it argues from *shared* premises.\n\n**Verdict: partially decomposable; residue = \"causal independence from human agency is itself value-conferring\"** — narrow, precise, contested, standing. *Confidence: medium-high.*\n\n### The pattern: a residue taxonomy\n\nRunning the six cases yields not one kind of bedrock but four recurring residue *types*:\n\n1. **Fittingness residues** (retribution; purity, on one of its three readings): claims of the form \"X makes response Y fitting,\" defended by the epistemic credentials of moral emotions, attacked by genealogical debunking ([Street's Darwinian Dilemma](https://philpapers.org/rec/STRADD) is the general template; Greene's dual-process debunking the psychological variant). Argument does not stop at these nodes; it changes register — from weighing to genealogy-versus-vindication.\n2. **Structural residues** (partiality): questions about the *logic* of normativity (do underived agent-relative reasons exist?), not about magnitudes.\n3. **Axiom-choice residues** (population ethics): formally mapped, provably jointly-inconsistent sets of precise principles; the disagreement is which axiom to surrender, and every candidate surrender is itself an argued position.\n4. **Permissive-zone residues** (risk temperament; Chang's parity generally): a characterized band of rationally permissible stances; argument targets the band's *boundaries*, and the within-band choice is a commitment rather than a truth-claim.\n\nTwo observations. First, **none of the six cases terminates in \"my weight versus your weight, nothing more to say.\"** Every residue is narrower than the original weighting, classifiable by type, and attackable in at least one further argumentative register. Second, **four of the six retain an irreducibly evaluative posit** (retribution, partiality, purity-on-one-reading, wilderness). So the strong dissolution claim is false, and the honest formulation is: *weightings decompose into structured, individually arguable claims, of which at most one or two per dispute are irreducible evaluative posits — and even those remain subject to genealogical, coherence, and boundary argumentation.*\n\nDoes this dissolve the objection that \"weightings survive full mutual understanding\"? Largely, by transformation. What survives is not a *weighting* — no scalar multiplier survived anywhere in the six cases — but a small set of typed, precise evaluative claims. [Fogelin's classic worry](https://philpapers.org/rec/FOGTLO-2) about deep disagreements (clashes of framework propositions that \"undercut the conditions essential to arguing\") still applies at the very bottom; but even Fogelin's own text names the countermove: \"surface these background propositions and then discuss them directly.\" Decomposition converts an amorphous deep disagreement into a *located* one. And empirically, most real disputes never reach the residue: the Liu & Ditto results and the fact-provision depolarization findings above indicate that layers 1–4 — especially compressed empirical beliefs — carry a large share of live disagreement variance.\n\n---\n\n## (d) Deliberus mechanics: a weighting-decomposition pass\n\nBrief and practical. The extraction pipeline already detects argument schemes and generates critical questions (CQs); weighing arguments would get one dedicated scheme plus a decomposition pass.\n\n**Scheme: Argument from Weighing.** Premises: *Option A promotes value V1; option B promotes value V2; V1 outweighs V2 in context C*; conclusion: *prefer A*. The weighting premise is auto-flagged as a candidate mother claim and receives:\n\n- **CQ1 — Covering value** (Chang): outweighs *with respect to what*? If no covering consideration can be named, the premise is not yet well-formed.\n- **CQ2 — Compression check**: is this weight a compressed empirical belief? Would the weigher endorse it under full information? Operational test: vary the stipulated facts — does the weight move? (Liu & Ditto predict it often will.)\n- **CQ3 — Elicitation stability** (Slovic; the MCDA bias literature): does the weight survive re-description, attribute splitting, and range changes? If not, mark it *constructed* rather than reported.\n- **CQ4 — Risk posture**: does the weight change when uncertainty is resolved or redistributed? If yes, split out the risk-function claim.\n- **CQ5 — Scope**: does the weight change when the set of affected parties changes? If yes, split out the moral-circle claim.\n- **CQ6 — Re-typing** (Baron & Spranca; Anderson): is this a scalar tradeoff at all — or a side-constraint, a lexical rule, or a mode-of-valuation claim? Protected-value markers (quantity insensitivity, outrage language, tradeoff denial) trigger re-typing rather than pricing.\n- **CQ7 — Specification** (Richardson): can the conflict be dissolved by adding where/when/unless clauses to the norms instead of weighing them?\n- **CQ8 — Residue typing**: if something remains after CQ1–7, classify it — fittingness / structural / axiom-choice / permissive-zone — and unlock type-appropriate follow-up schemes (genealogical-debunking-plus-vindication pairs for fittingness; impossibility mapping for axiom sets; explicit parity marking with commitment semantics for permissive zones).\n\nIn QBAF terms, residue nodes are a distinguished species of sorry marker — *evaluative posits* — whose strength propagates transitively to the mother claim like any other decomposition, so a reader who attacks or supports the residue moves the original weighting's badge.\n\n---\n\n## Per-case verdicts\n\n| Case | Verdict | Residue, named precisely | Confidence |\n|---|---|---|---|\n| 1. Retribution | Hits a bedrock-shaped residue | \"Culpable wrongdoing makes the wrongdoer's suffering non-instrumentally fitting\" | High |\n| 2. Partiality | Partially decomposable; the magnitude fully dissolves | Existence of underived agent-relative reasons — a structural question, not a quantum of love | High |\n| 3. Population ethics | Fully decomposable into axioms | Axiom choice under proven joint inconsistency (Arrhenius) | High |\n| 4. Purity/sanctity | Partially decomposable; the residue's own type is unresolved | Fittingness posit *vs.* debunkable affect program *vs.* implicit harm belief — a live three-way dispute | Medium |\n| 5. Risk temperament | Fully decomposable, with a formal residue | Choice of risk function r(p) within a contested rationally-permissible band (Buchak) | High |\n| 6. The untouched | Partially decomposable | \"Causal independence from human agency is itself value-conferring\" (Elliot's genesis claim, narrowed) | Medium-high |\n\n---\n\n## Implications for Deliberus\n\n**The move is sound as reframing, not as elimination.** The platform can claim — and defend with the sources above — that no weighting is an argumentative dead end: every weighting premise decomposes into arguable structure, and what cannot be further decomposed can be precisely typed and argued in a different register. It should *not* claim that decomposition eliminates evaluative bedrock; four of six test cases retain an evaluative posit, and overclaiming here would be both false and easily punctured.\n\n**Population ethics is the existence proof.** Five decades of philosophical decomposition turned \"creating people just matters less — that's how I weigh it\" into named principles, conditional-reasons analyses, and impossibility theorems. That is the mature end-state a weighting-decomposition pass points toward: not consensus, but disagreement so precisely located that its residue has a theorem attached.\n\n**Most convergence value lives in layers 1–4.** The compressed-empirical-belief layer alone (Liu & Ditto; the fact-provision depolarization results) suggests a large fraction of practical weighting disputes will move under decomposition without ever touching a residue. The wilderness case shows a second bridging mechanism: decomposition re-derives the contested value from premises both sides already share (authenticity in art and achievement), shrinking the gap even when a residue stands.\n\n**The taboo-tradeoff adoption risk is real and must shape the UX.** The Tetlock–Fiske–Ginges literature warns that for genuinely sacred values, *the act of decomposition itself* can be experienced as desecration: merely contemplating a taboo tradeoff triggers outrage and cleansing in observers, and sweetening analysis with instrumental stakes *backfires*. Three design consequences. (1) Never present scalar-pricing prompts (\"how many X is one Y worth?\"); CQ6 exists precisely to route protected values into re-typing instead of pricing. (2) Frame decomposition in specification language — \"what does honoring this value require here?\" — which lands in the *tragic/routine* frame people engage with rather than the *taboo* frame they punish; the reframing results (Tetlock 2003) show this door is genuinely open. (3) Make weighting-decomposition an invitation, not an ambush: the sorry-marker idiom (an open door, not a verdict) is already the right register, and acknowledgment of the value should precede analysis of it — the symbolic-concession findings suggest recognition is a precondition for engagement, not a nicety. The mere-contemplation effect adds a UI requirement: keep \"analyzing this argument\" visibly distinct from \"endorsing this tradeoff,\" because observers judge deliberators, not just conclusions.\n\n**Parity deserves first-class representation.** Chang-style parity and Buchak-style permissive bands are honest terminal states that are *not* disagreements about truth. A graph that can mark a node \"decomposed to parity: both commitments rationally permissible\" gives users a principled way to stop arguing that is neither victory nor stalemate — a novel product primitive, and a truthful one.\n\n---\n\n## Strongest objection to my own conclusions\n\nThe relocation worry, generalized: **the decomposition schema is not value-neutral, and pretending otherwise is a worldview imposition.** Decomposing a weighting into empirical beliefs + risk posture + scope + aggregation quietly presupposes a broadly consequentialist bookkeeping in which values are the kind of thing that *have* scopes and aggregation rules. The partialist who insists that \"she is my daughter\" is a complete reason, the Williams-style critic who says thick concepts do not factor, and Anderson's pluralist about modes of valuation all deny the schema's presupposition — and the taboo-tradeoff findings can be re-read not as an adoption obstacle but as ordinary people *correctly detecting* that the analytic frame smuggles in a metaethic. On this reading, the four residue types are partly artifacts of forcing every value through one sieve; a different sieve would find different residues, and the choice of sieve is itself the deep disagreement. The available response — treat the decomposition schema as an arguable claim in the graph, offer multiple decomposition frames in the spirit of worldview lenses, and represent the schema-choice dispute rather than presupposing its answer — is coherent, but it concedes real ground: decomposition passes then inherit the platform's hardest open problem (worldview-relative structuring) rather than bypassing it. Two smaller concessions compound this. If Dancy's holism is right, decomposed parts lack stable cross-context contributions, so accumulated general lemmas will overstate their transferability. And decomposition can *multiply* visible disagreements rather than reduce them — a more precisely located disagreement is only progress if locatedness is itself valuable, which is a commitment of this platform, not a theorem.\n\n---\n\n*Research conducted 2026-07-05 by an AI agent (web-sourced). Citations link to primary or academic sources where available; quotations should be verified against the linked originals before being relied upon.*"}