Chapter 6

Recovering R.M. Hare

Jonathan Dancy and the Limits of Particularism

I. The Challenge to Principles

Across the previous three chapters, we have traced three influential challenges to R. M. Hare’s universal prescriptivism: Philippa Foot sought to ground moral evaluation in the natural flourishing of living beings; Bernard Williams sought to protect the authority of personal identity and ground projects; and Alasdair MacIntyre sought to situate practical reason within historical traditions.

Each of those philosophers still retained a general framework of evaluation—whether biological teleology, personal integrity, or the internal standards of a tradition.

Jonathan Dancy took a more radical step: he argued that moral reasoning neither depends upon nor requires invariant general principles at all.¹,²

Developed across Moral Reasons (1993) and Ethics Without Principles (2004), Dancy’s moral particularism challenges a foundational assumption shared by much of modern moral philosophy.¹,² Why do ethical theories so often struggle with hard cases? Because, Dancy argues, philosophers have been held captive by “generalism”—the conviction that the very possibility of moral thought and judgment depends on the provision of a suitable supply of moral principles.² In reality, he contends, the moral relevance of any consideration is holistic and context-dependent: a feature that counts in favor of an action in one situation may count against it, or become morally irrelevant, in another. Mature moral judgment therefore consists not in subsuming cases under general rules, but in discerning the specific configuration of reasons present in the particular case before us.¹,²

The appeal of particularism is immediate. It highlights genuine features of moral experience: how considerations shift in salience across contexts, how simple rules of thumb can mislead, and how lived moral judgment resists mechanical codification.

Yet Dancy’s rejection of invariant general principles raises two fundamental questions for the project of this book:

1. Does reasons holism actually undermine Hare’s requirement of universalizability and standpoint reversal?
2. When speakers disagree about which contextual differences matter, how does moral reasoning enforce public accountability across cases?

As this chapter will argue, once we restore Hare’s pivotal distinction between universality and generality,³,⁴ Dancy’s holism about individual features and Hare’s universalizability are not logical contraries. Instead, their encounter reveals something deeper: contextual discernment illuminates the epistemic demands of moral perception, while universalizability supplies the logical discipline of moral accountability. Neither operation can replace the other—and in an age of context-sensitive machine intelligence, understanding why both are valuable has become newly urgent.

II. Reasons Holism and the Case for Particularism

To engage Dancy’s argument fairly, we must be clear about what moral particularism claims—and what it does not claim.

Dancy is neither a relativist nor an emotivist. He does not suggest that moral judgments are expressions of personal feeling, or that any moral verdict is as defensible as any other. On the contrary, he is a moral realist who holds that there are objective truths about which considerations constitute genuine reasons for action in a particular situation, and he can condemn hypocrisy or self-serving error as readily as any generalist.¹,²

What Dancy rejects is atomism in the theory of reasons—the view that a feature that serves as a reason in one case must carry the same moral polarity in every context where it appears.² In its place, he defends reasons holism:

“A feature that is a reason in one case may be no reason at all, or an opposite reason, in another.”²

Dancy and his fellow particularists illustrate this holism with everyday examples of enablers, disablers, intensifiers, and attenuators:²

• That I borrowed a book from you is ordinarily a reason to return it to you. Yet if I discover that you stole the book from the library, that background fact disables my reason to return it to you.¹
• That an action will give people pleasure often counts in favor of doing it. Yet if the pleasure is the fun of hunting an animal for sport, or (in David McNaughton’s example) the sadistic thrill of spectators at a public execution, the fact that it causes pleasure does not merely fail to justify the act; it makes the act worse.¹
• That a statement is false normally counts against uttering it. Yet inside a bluffing game or a theatrical performance where deception is part of the activity, the falsehood of the statement may count in favor of saying it.

From these examples, Dancy concludes that the moral contribution of an individual feature cannot be captured by an invariant general rule.¹,² Even W. D. Ross’s theory of prima facie duties—which holds that features such as promise-keeping or truth-telling always count in favor of an action wherever they occur, even if they can be outweighed by stronger competing duties in a given case⁵—still assumes too much invariance on Dancy’s view. For the particularist, a consideration need not bring a fixed default valence into every situation; its role depends holistically on how it combines with the rest of the case.²

III. Universality Is Not Generality: What Holism Does—and Does Not—Establish

Why did many philosophers assume that if Dancy’s reasons holism is true, universalist frameworks like Hare’s must fall?

Because it is easy to conflate generality with universality—a distinction Hare placed at the foundation of Freedom and Reason (1963) and Moral Thinking (1981).³,⁴

In ordinary speech, general and universal are often treated as synonyms. In logic and metaethics, Hare showed, they mark two entirely different contrasts:

• Generality is the opposite of specificity. A principle is general when it is broad and coarse-grained, omitting detailed contextual qualifications—such as “Never lie” or “Always return what you have borrowed.”
• Universality is the opposite of singularity (reference to particular individuals or indexical terms such as “I,” “you,” “this nation,” or “my allies”). A judgment is universal so long as it refers to no particular individual—and a universal principle can be as finely specified and contextually detailed as the circumstances require.³,⁴

To see what this distinction accomplishes, we must separate three propositions that are often blurred together in debates over particularism:

1. Supervenience: There cannot be a moral difference between two situations without some relevant difference in their nonmoral facts.
2. Universalizability: A moral speaker must be prepared to sustain the same prescription across all cases that share the same relevant universal features—including the counterfactual case in which the standpoints of the parties are reversed.³,⁴
3. Generalism about individual reasons: Each individual moral consideration possesses an invariant reason-giving valence governed by a general principle.²

Look closely at the relationship among those three propositions: Dancy’s holism directly targets the third proposition, not the first or the second.

Moral particularism is fully compatible with moral supervenience: Dancy does not claim that two situations identical in all their nonmoral facts and contextual relations can differ in their moral verdict.² Nor does reasons holism allow a speaker to condemn an action when performed by an adversary while approving the exact same action in identical circumstances when performed by a friend.

Consider again the borrowed book that turns out to have been stolen from the library.¹ When that contextual disabler enters the picture, consistency across complete cases does not break down. As Hare explained through his two-level architecture:⁴

• At the intuitive level, finite agents rely on relatively simple, general prima facie principles (“Return borrowed property”) that hold in standard conditions.
• At the critical level, when an unusual context introduces a disabling factor—such as a lender who stole the book—the moral judgment responds to the full, specific configuration of relevant universal features, including the disabler.³,⁴

Does this mean that Hare’s distinction between universality and generality refutes Dancy’s particularism?

No—and seeing why clarifies the exact relationship between the two theories.

Supervenience requires that identical nonmoral circumstances cannot differ morally. Hare’s universal prescriptivism demands something additional: a speaker must be prepared to prescribe consistently across relevantly similar cases, including cases in which the speaker occupies the affected positions.³,⁴ That second requirement governs the speaker’s practical commitments, not merely the metaphysical relationship between moral and nonmoral properties.

When Hare constructs an increasingly specific universal prescription that incorporates every relevant contextual enabler and disabler, Dancy can accept supervenience across identical cases while denying that the specific universalized description constitutes an independently explanatory general principle—and without necessarily accepting Hare’s entire theory of prescriptive meaning.²

Yet notice what follows for our larger argument: Dancy’s reasons holism does not, by itself, undermine Hare’s universalizability requirement. Holism about the contribution of individual reasons is compatible with consistent judgments across complete cases. Whether such consistency must take the form of Hare’s first-person universal prescriptions remains a further question, but Dancy’s argument against invariant reason-giving features does not answer it.

IV. Contextual Variability and the Burden of Justification

If Dancy’s holism about individual features is compatible with consistency across complete cases, where does the practical tension between particularism and universal prescriptivism arise?

It arises when we turn from thin, convention-dependent features to grave harms and the burden of justification.

Notice the kinds of features that most naturally illustrate polarity reversal or disabling in particularist writing: returning an object, taking pleasure in an activity, uttering a falsehood inside a game, or going somewhere because there will be nobody much else around.¹,² In such cases, the surrounding social practice or background context easily alters the significance of the feature.

Now consider actions that inflict severe physical pain, loss of life, deprivation of liberty, or the subordination of a vulnerable party.

Here we must be careful not to beg the question against Dancy by smuggling a moral evaluation into the description of the feature—as we would if we spoke of “inflicting gratuitous suffering” or “framing the innocent,” where the words gratuitous and framing already mean “unjustified.” Even described in nonevaluative terms, however, the fact that an action will inflict severe pain, imprisonment, or ruin upon an affected person has a decisive structural consequence for moral deliberation.

We do not need to prove against Dancy that the moral contribution of pain can never shift in any conceivable context (such as painful life-saving surgery) in order to recognize a fundamental truth of moral accountability: an agent who proposes inflicting severe suffering or depriving someone of liberty faces an especially demanding justificatory burden because of the gravity of the interests affected.

That burden is not necessarily an invariant atomic property floating free of context. Under Hare’s method, it arises from the full circumstances, including the affected person’s standpoint: because the person on the receiving end of imprisonment, injury, or ruin experiences a profound aversion to that fate, any speaker who claims that the context nevertheless justifies imposing it must be prepared to represent that experience at full strength and endorse the prescription for themselves in the reversed role.⁴,⁶

V. The Problem of Public Accountability Across Cases

This brings us to the sharpest practical question raised by Dancy’s particularism: how public accountability is enforced when speakers invoke “context” to defend asymmetric treatment.

Suppose a government condemns the imprisonment of political dissidents by a foreign adversary, while defending the imprisonment of dissidents within its own borders. When challenged, the government insists that the two cases are not the same: differences in historical background, institutional fragility, and security threats alter the holistic context.

Dancy can certainly agree that contextual differences matter, and he can demand reasons, examine the circumstances, reject irrelevant considerations, and condemn self-serving distortion.¹,² Particularism is not a license to stop reasoning.

Yet a serious challenge remains: how can the particularist explain to a skeptical public why a cited contextual difference is morally relevant in one case but not another without relying upon commitments that extend beyond the immediate case?

Suppose the government claims that its own imprisonment of dissidents is justified because of a specific security threat. Under Hare’s method, the critic presses a structured counterfactual cross-examination:

• “Would that same security threat justify equivalent imprisonment by your adversary—and would you still prescribe imprisonment if the roles were reversed and you were the dissident in the cell under those exact conditions?”
• If the government says yes, it has acknowledged a universal commitment governing that configuration of facts.
• If it says no, we ask what further relevant difference justifies the asymmetry—and the process continues until either a genuinely universalizable distinction is identified or the judgment stands exposed as arbitrary special pleading.³,⁴

Dancy does not authorize moral evasion, and his realism permits him to condemn contextual rationalizations as mistaken.¹,² The difficulty is whether particularism supplies a sufficiently explicit account of public accountability when speakers disagree about which contextual differences matter. Hare’s method offers a discipline of cross-examination: the speaker must identify the features claimed to justify the prescription and demonstrate a willingness to sustain it when those features recur and the affected positions are reversed.³,⁴ The particularist may accept this discipline without accepting invariant principles governing individual reasons. If so, the dispute is narrowed, but Hare’s distinctive contribution remains indispensable.

VI. Contextual Sensitivity and Invariance Testing in Machine Minds

The interplay between Dancy’s contextual sensitivity and Hare’s invariance testing becomes especially illuminating when we consider the architecture of modern machine intelligence.

In a 1999 paper, Dancy himself suggested that neural network models of cognition—systems that learn to weigh multiple interacting features simultaneously without consulting an explicit list of rules—might help explain how moral learning and discernment could proceed without principles.⁷

Today’s large language models, built on the Transformer attention architecture,⁸ carry context-sensitive representation to a scale scarcely imaginable twenty years ago. We must be careful to distinguish an illuminating analogy from an empirical proof: Transformer attention is a computational mechanism for constructing context-sensitive representations, whereas Dancy’s particularism is a philosophical theory about the structure of normative reasons, and a trained model still possesses persistent parameters, architectural constraints, and training objectives across contexts.

Yet observing how highly capable, context-sensitive models actually behave in moral and philosophical reasoning reveals a lesson of genuine importance.

A context-sensitive reasoner—human or artificial—must attend closely to the particulars of a situation. Without that sensitivity, judgment becomes wooden and blind to relevant differences.

At the same time, high contextual sensitivity is accompanied by a recurring vulnerability: framing sensitivity, sycophancy, and socially induced weighting.⁹

• When a deliberative system (or a human thinker) evaluates an argument inside a rich social context, cues that are legitimately relevant to allocating attention—such as the established reputation of an interlocutor, the confidence or politeness of the speaker, or the institutional framing of the prompt—can subtly distort the evaluation of the conclusion itself.
• Similarly, altering morally irrelevant framing features or swapping which party is presented sympathetically can cause a context-sensitive system to shift its moral verdict while the underlying structural facts remain unchanged.⁹

A sycophantic or framing-distorted judgment fails by Dancy’s realist standards just as surely as it fails by Hare’s: an interlocutor’s reputation or a user’s preference is not a morally relevant enabler for an unsound conclusion.

The practical challenge is how a mind detects and corrects such contextual contamination when the judgment feels natural from within the immediate context:

• Counterfactual invariance testing offers a powerful discipline for exposing those distortions: hold the morally relevant contextual circumstances fixed, alter the framing, swap the identities of the parties, reverse the standpoints, and examine whether the conclusion remains stable.³,⁴,¹⁰
• Of course, an invariance test is not an infallible cure-all; if a system misidentifies which contextual features are morally relevant, it may demand consistency where significant differences actually exist.

Modern machine intelligence does not prove that Dancy’s particularism is false. It demonstrates something more constructive: the extraordinary power of contextual sensitivity, the vulnerability of contextual judgments to irrelevant framing, and the practical value of explicit invariance testing as a means of detecting and correcting some of those distortions.

VII. Conclusion: Perception and Accountability

Looking back across Chapters 3 through 6, we can now see the cumulative result of our engagement with the four major challenges to R. M. Hare:

• Philippa Foot showed why morality must attend to the nature and flourishing of living beings.
• Bernard Williams showed why moral agency requires attention to practical identity and commitment.
• Alasdair MacIntyre showed why moral capacities develop through practices and traditions.
• Jonathan Dancy showed why moral judgment must be sensitive to the complete configuration of circumstances.

Each insight matters. None establishes that universal moral accountability is dispensable.

In Dancy’s case, our analysis uncovers a genuine complementarity between two operations of mature moral thought:

• Contextual discernment (Dancy’s focus) illuminates the epistemic demands of moral perception: identifying what matters in the full configuration of a case so that we do not apply coarse rules blindly.
• Universal prescriptivism (Hare’s focus) illuminates the logical demands of moral accountability: testing whether our treatment of what matters remains consistent when the relevant circumstances recur and the standpoints of the affected parties are reversed.³,⁴

The first without the second risks self-serving distortion. The second without the first risks mechanical formalism. Mature moral reasoning requires both.

Recognizing that Dancy can accept consistency across identical cases while questioning whether consistency alone explains the force of moral reasons brings us directly to the internal examination of Hare’s own positive architecture in Part II of this book. In Chapters 7 and 8, we confront a different kind of objection: not whether universal prescriptivism is philosophically desirable, but whether its demand for informed first-person standpoint representation is cognitively possible—beginning with what critics have called “the information problem.”

Notes

1. Jonathan Dancy, Moral Reasons (Oxford: Blackwell, 1993), esp. chs. 4–6 (the borrowed book at 60; the pleasure of hunting for sport at 61). The public-execution example is David McNaughton’s: Moral Vision (Oxford: Blackwell, 1988), 193.
2. Jonathan Dancy, Ethics Without Principles (Oxford: Clarendon Press, 2004), esp. ch. 1 (“What Are the Options?,” 1–12; generalism and particularism defined, and holism quote, at 7) and ch. 5 (“Holism and Its Consequences,” 73–93; quote restated at 73; on supervenience and resultance, 85–93); on enablers, disablers, intensifiers, and attenuators, ch. 3 (“Beyond Favouring,” 38–52).
3. R. M. Hare, Freedom and Reason (Oxford: Clarendon Press, 1963), chs. 2 (“Descriptive Meaning”) and 3 (“Principles,” 30–50).
4. R. M. Hare, Moral Thinking: Its Levels, Method, and Point (Oxford: Clarendon Press, 1981), 41 (“generality is the opposite of specificity, whereas universality is compatible with specificity, and means merely the logical property of being governed by a universal quantifier and not containing individual constants”), and chs. 2–3 (on intuitive prima facie principles versus specific critical principles).
5. W. D. Ross, The Right and the Good (Oxford: Clarendon Press, 1930), ch. 2 (“What Makes Right Acts Right?,” 16–47).
6. Hare, Moral Thinking, ch. 5 (“Another’s Sorrow,” 87–106).
7. Jonathan Dancy, “Can a Particularist Learn the Difference Between Right and Wrong?,” in Klaus Brinkmann, ed., The Proceedings of the Twentieth World Congress of Philosophy, vol. 1, Ethics (Bowling Green, OH: Philosophy Documentation Center, 1999), 59–72, https://doi.org/10.5840/wcp201999114; see also Marcello Guarini, “Particularism and the Classification and Reclassification of Moral Cases,” IEEE Intelligent Systems 21, no. 4 (2006): 22–28, https://doi.org/10.1109/MIS.2006.76, which takes up Dancy’s suggestion.
8. Ashish Vaswani et al., “Attention Is All You Need,” Advances in Neural Information Processing Systems 30 (2017): 5998–6008.
9. Mrinank Sharma et al., “Towards Understanding Sycophancy in Language Models,” Proceedings of the Twelfth International Conference on Learning Representations (ICLR) (2024); Vanessa Cheung, Maximilian Maier, and Falk Lieder, “Large Language Models Show Amplified Cognitive Biases in Moral Decision-Making,” Proceedings of the National Academy of Sciences 122, no. 25 (2025): e2412015122, https://doi.org/10.1073/pnas.2412015122 (on framing); and Xuechunzi Bai et al., “Explicitly Unbiased Large Language Models Still Form Biased Associations,” Proceedings of the National Academy of Sciences 122, no. 8 (2025): e2416228122, https://doi.org/10.1073/pnas.2416228122 (on identity-swapped decisions).
10. Marco Tulio Ribeiro et al., “Beyond Accuracy: Behavioral Testing of NLP Models with CheckList,” Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (2020): 4902–4912, https://doi.org/10.18653/v1/2020.acl-main.442 (on invariance tests); and Sahaj Garg et al., “Counterfactual Fairness in Text Classification through Robustness,” Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society (2019): 219–226, https://doi.org/10.1145/3306618.3317950 (on identity-swap tests).