Chapter 2

Recovering R.M. Hare

The Archangel Is the Test

I. The Work of the Ideal

A person can take promises seriously and still face a situation in which keeping a promise would expose someone to grave harm. The conflict arises because both claims matter. Repeating the importance of fidelity does not resolve it; neither does pointing out that harm matters too. Moral judgment has reached a question that the principles guiding ordinary conduct cannot answer by themselves.

Hare introduced the Archangel to clarify what answering such a question requires. The ideal reasoner knows the relevant facts, understands the positions of those affected, and can follow the implications of a prescription without the distortions that obstruct human deliberation. By distinguishing this critical thinking from the intuitive thinking through which we usually act, Hare explains both how principles are justified and why finite people need them.¹

An objection based on the Archangel’s superhuman capacities therefore has to identify what those capacities are doing in the theory. If an account required parents to complete an exhaustive calculation before comforting their children, its practical demands would deserve criticism. Hare’s account explains why parents should possess the affection and dispositions that move them directly. Their limitations enter the justification of those dispositions.

The ideal also makes a demand on anyone who claims that a principle should govern other people. We need to ask what the principle prescribes when its implications are understood from the positions it affects. A claim that survives because the speaker overlooks someone else’s predicament has acquired no authority from that oversight. Better knowledge should expose the error. A theory that explains why it does so has given us a reason to seek the knowledge.

The Archangel’s importance lies in this relation between justification and practice. Critical reasoning supplies the standard by which conduct and the principles guiding it are assessed. Our limitations help determine how we should pursue that standard in ordinary life. Treating the limitations as an objection to the standard leaves the original conflict where we found it.

II. What Critical Thinking Decides

In Moral Thinking, Hare imagines two limiting figures. The Archangel can deliberate critically without the ignorance, confusion, and weakness that afflict human reasoners. The pure prole depends entirely on acquired principles and dispositions and is incapable of critical thinking even when there is time for it. Hare explicitly refuses to divide humanity into these two classes. Actual people share their characteristics in different degrees and circumstances.¹

The contrast identifies two functions. Critical thinking determines what ought to be done and selects principles for agents who cannot deliberate ideally on every occasion. Intuitive thinking enables those agents to act through the principles and dispositions they have acquired. Their relationship is central to the theory: the reliability of ordinary moral conduct depends partly on the quality of the principles that critical reflection has selected for it.²

The Archangel’s task includes the complete method of justification. He can examine the available actions, represent their consequences for the people affected, and reach a prescription. Exposing an arbitrary exemption is one part of this work. The argument also explains how interests that conflict are to enter a common deliberation.

Representing another person’s situation requires more than describing it from a safe distance. The deliberator must grasp what occupying that position would mean, with its preferences and their strength. Hare argues that adequate first-person representation brings a corresponding conditional preference into present deliberation. Universalization requires equal weight for preferences of equal strength across the affected positions. The resulting method is preference utilitarianism.³

The deliberation must weigh the preferences of all affected people together. Showing that one person’s gain exceeds any single other person’s loss does not show that it exceeds their combined losses. Chapters 7 and 8 examine the argument connecting representation, impartial weighting, and aggregation.

Hare distinguishes this method from the adoption of an independent Principle of Utility. His route to the conclusion proceeds through the requirements of moral judgment and the preferences of the people concerned. The Archangel embodies reasoning conducted under those requirements; defining a wise and impartial creature would not by itself establish them.⁴

This matters when interests conflict. A decision can be justified while leaving someone with a preference that cannot be satisfied. The disappointment remains real. What has been settled is what to do in the presence of that conflict. To insist that critical thinking has failed because it has not made everyone want the same thing would confuse agreement in judgment with the disappearance of competing interests.

The same distinction governs the treatment of fanaticism. A willingness to endure one’s own proposed persecution does not establish that everyone else’s position has been adequately represented. Hare’s mature discussion distinguishes the purported fully critical thinker who reaches a counterutilitarian conclusion from the person whose extraordinary preferences make an apparently fanatical choice itself utilitarian. Hare argues that the first is excluded if the argument of Moral Thinking, Chapter 5, is correct; the second does not contradict that argument.⁵ The Archangel cannot be reduced to a test that registers the fanatic’s willingness to suffer and then stops.

Nor does universality require a judgment to ignore context. A promise, a relationship of dependence, or a responsibility to a child can distinguish one case from another. The demand is that the judgment survive an exchange of the people occupying the relevant positions, with those circumstances faithfully preserved. A prescription about parents and children can recognize the relationship while applying to anyone who occupies either role.

The authority of the ideal comes from the reasoning it represents. Calling someone an Archangel would confer no authority on their pronouncements. Their conclusions would still have to follow from the facts and the argument. The thought experiment makes those requirements explicit and supplies a standard against which actual reasoning can be examined.

III. Principles for Finite Agents

The question of how people should live includes the fact that they have limited time, imperfect information, and interests that can distort their judgment. These are facts about the agents for whom moral guidance is needed. Critical thinking must take them into account when assessing the principles they should learn and the dispositions they should cultivate.

A person who keeps promises only after reconsidering the advantage of each promise will give others little reason to rely on them. The capacity to rely on another person depends on something more durable: a disposition that continues to operate when performance becomes inconvenient. Such a disposition shapes what the person notices, what they feel, and what they are prepared to do. Its importance cannot be captured by treating it as a convenient slogan available for occasional use.

Hare gives intuitive principles this depth. They belong to character and carry motivations and feelings with them. Critical thinking assesses the effects of cultivating and accepting them in the circumstances of actual life. The assessment includes the mistakes people predictably make, the temptations they face, and the consequences of making exceptions too easily.⁶

Consider a community deciding how to educate its children about truthfulness. It must consider what happens when people acquire a settled reluctance to deceive, what happens when that reluctance is weak, and how children will learn to recognize circumstances in which telling the truth would itself betray someone who needs protection. Teaching a formula and leaving every occasion to improvisation would neglect much of the problem. So would making the formula immune to examination.

A critically justified education can cultivate a strong disposition to tell the truth together with the ability to understand its justification. A child can come to see why others need reliable information and why convenient deception damages their lives. When a difficult case arises, that understanding gives the child something with which to reason. The moral education has equipped the person both to act and to question responsibly.

The same structure applies to affection. A parent can respond immediately to a child’s distress because the child matters to them. Critical justification concerns the responsibilities and dispositions through which that response becomes dependable, including the costs of replacing direct concern with constant deliberation. The parent does not need to rehearse that justification whenever the child cries.

The intuitive level therefore has a substantial place in moral life. The critical level explains why it should have that place and which commitments deserve cultivation. A settled disposition to care for others and a settled disposition to obey a cruel authority may both be psychologically effective. Their effectiveness alone does not tell us which we should teach.

This is the question an appeal to inherited conviction cannot settle. What we happen to have learned provides material for examination. It cannot, merely by being deeply learned, determine what we ought to pass on. Critical thinking is necessary at the point where the continuation of a moral culture requires a judgment about its content.

IV. The Danger of the Fallen Archangel

Hare’s warning about “fallen archangels” concerns the confidence with which people mistake their actual powers of thought and character for the capacities of the ideal. A person may be clever enough to produce an elaborate justification and still be poor at recognizing the advantage it protects. The danger becomes acute when they are tempted to set aside a principle that normally restrains them.⁷

Imagine someone who has made an inconvenient promise. They discover that breaking it would release time for an activity they now regard as more beneficial. Their calculation describes that activity in generous detail while treating the disappointed person’s expectations as a minor inconvenience. The calculation’s fluency conceals a failure of representation. The promisee’s position has been discounted at the stage when the case was assembled.

The critical response begins by correcting that case. What did the promise enable the other person to plan? What alternatives have they forgone? What would the proposed exception mean if it were invoked against the deliberator in corresponding circumstances? A faithful assessment may still justify departing from a promise. It must do so with the promise and its effects fully represented.

The warning also bears on the decision to calculate at all. Someone who repeatedly discovers exceptional reasons to disregard the same obligation has reason to examine the reliability of that discovery. A disposition to reopen settled questions whenever compliance becomes costly can itself be assessed. Critical reflection can justify resistance to one’s own improvised calculations.

But circumstances can make further deliberation necessary. If two commitments cannot both be fulfilled, simply insisting on both will not produce an action that honors them. Hare recognizes that some conflicts can be managed through familiar priorities. Others require examination of the particular case, revision of a principle, or both. His warning about the danger of deliberating under pressure accompanies an explanation of why we sometimes have to do it.⁷

The two levels are therefore functions within moral thought, not sealed compartments between which a person must switch without overlap. Someone can feel the force of a commitment while examining whether it should govern the present decision. The feeling may persist after the decision. Its persistence shows that the commitment remains part of the person’s character; it does not establish that the final judgment is contradictory.⁸

There is practical work to do before such a conflict occurs. People can investigate recurring problems, examine the reasons behind their rules, and seek criticism from those likely to notice what they overlook. Institutions can make the relevant perspectives harder to exclude by requiring evidence and an answer to opposing arguments. These arrangements have to be judged by how well they improve deliberation. Their existence does not confer immunity on their decisions.

Human fallibility thus matters throughout the account. It shapes the principles we should cultivate, the occasions on which we should reconsider them, and the precautions appropriate to that reconsideration. The Archangel makes clear what the precautions are intended to secure.

V. Commitment Under Reflection

The objection from alienation challenges the relation between these two levels. Can a person inhabit a commitment wholeheartedly while understanding its place in a critical justification? Or does reflection turn an attachment into an instrument and thereby destroy what made it an attachment?

Michael Stocker’s hospital visitor gives the concern a memorable form. The visitor comes because visiting the sick is a duty, while lacking the direct concern for the friend that would make the visit an expression of friendship. Stocker’s broader argument challenges theories whose ultimate justifications cannot be embodied in the motives required by valuable relationships. He considers indirect strategies as well, so a response must explain how the relation between motive and justification can be lived without self-deception.⁹

The failure in the hospital is the absence of concern for the friend. Hare’s account requires us to ask why a critically justified disposition could not include precisely that concern. A person can want a friend’s recovery for the friend’s sake and understand why a life containing such attachments is worth sustaining. The second thought need not displace the first. To establish the objection, the critic must show why understanding that justification defeats, distorts, or makes self-deceptive the direct concern that the relationship requires.

Bernard Williams presses the objection against Hare’s two-level structure. The concern reaches beyond the difficulty of changing one’s focus. Ethical commitments shape how their possessor sees a situation, while the utilitarian assessment appears to regard those commitments as dispositions to be maintained for their beneficial effects. Williams argues that the two understandings cannot remain stable under the thorough reflection Hare’s theory itself encourages.¹⁰

The reply turns on the relation between the object of a commitment and the justification of cultivating it. A friend’s recovery can be the direct end of one’s concern while that concern remains open to impartial assessment. Wanting this person to recover does not itself assert a theory about what ultimately justifies moral conduct. Critical assessment includes the recovery and the relationship through the interests and preferences of those affected.⁴ The further question concerns how that concern may be expressed when other people’s positions are equally represented. An argument is still needed to show why understanding this justification must convert direct concern into a pretense.

The distinction also clarifies what an intuitive principle commits its holder to. A serious commitment to keeping promises need not contain the judgment that every promise must be performed regardless of all other circumstances. It can remain a strong prescriptive commitment even when it is overridden in a particular case. Hare makes this distinction explicit: overriding a prima facie principle does not require ceasing to hold it.¹¹

Hare also directly answers the charge that reflection erodes commitment. In his 1992 discussion of moral education, he considers children learning to resolve a conflict between promises and to refine their principles through experience. He challenges Williams’s description of such development as erosion and presents the cultivation of critical competence as part of moral maturation.¹²

The example exposes a distinction the objection needs to respect. A person can become less mechanical in applying a principle while becoming more trustworthy in the treatment of others. They may learn to recognize a competing responsibility, to distinguish a serious conflict from an excuse, and to explain a necessary departure to the person whose expectations they cannot meet. None of this requires indifference to keeping their word. It requires understanding what keeping one’s word means in a life that contains other people and other responsibilities.

Love presents a further issue because it concerns a particular person, rather than simply adherence to a general rule. Here critical assessment must represent the particular relationship accurately. It must represent what the relationship means to those who have it, the good they find in it, and the losses its destruction would cause. A description that replaces the beloved with an interchangeable source of benefits has omitted facts that matter to the people whose lives are being assessed.¹³

A parent who accepts responsibilities toward their own child can also recognize corresponding responsibilities in other parents. They can care directly for this child while understanding why children should be able to rely on those who care for them. That understanding does not require a hidden belief that their attachment is fraudulent, or that the child matters only while no marginally more attractive use of the parent’s attention presents itself.

What requires justification is the treatment that follows from the attachment when other lives are at stake. A parent’s love can explain the urgency of protecting a child without settling whether the parent may sacrifice an innocent stranger to do so. The stranger’s position has to enter the judgment with its own force. Giving it that place challenges an exemption the parent might want; it does not establish an incompatibility between love and reason.

There can be anguish in such a judgment. A person may continue to feel the pull of what they cannot rightly do. Hare’s account gives those feelings a place within moral experience while distinguishing them from the conclusion reached through critical deliberation. The objection must show where that account produces incompatible commitments or defeats the relationships it justifies. The presence of painful reflection does not supply the demonstration.

The distinction matters beyond the defense of Hare. A commitment’s psychological depth cannot establish its moral authority. People can be deeply attached to arrangements that damage others. An account that protects their capacity for wholehearted commitment must still explain how the resulting conduct can be challenged. Hare’s two-level structure preserves that question within the account of character itself.

VI. Correction Without Omniscience

The practical force of the Archangel becomes clearer when we consider an argument that exposes an error without possessing ideal knowledge. Abraham Lincoln’s notes on slavery, conventionally dated to 1854, examine successive claims about what might entitle one person to enslave another. If lighter skin confers the right, a still lighter person can invoke it against the advocate. If superior intellect confers it, the advocate is subject to anyone intellectually superior. If self-interest supplies the justification, the same justification becomes available to someone whose interests would be served by enslaving the advocate.¹⁴

The argument makes the defender answer for the general ground they have offered. Its force comes from applying that ground beyond the case in which it protects their own advantage. A reason that changes its authority when its advocate becomes the person subjected to it has disclosed an exception that needs justification.

Hare develops this discipline through the full representation of affected positions. An imagined exchange must include the loss of freedom, the relationships and projects disrupted, and the preferences of the person who would actually bear those burdens. The person defending a practice cannot substitute a contented beneficiary of it for someone whose opposition is among the relevant facts.

We can recognize a failure of this kind without possessing every fact an ideal deliberator would know. A decision may omit a person whose interests it plainly affects. An alleged benefit may depend on a factual claim that can be corrected. A difference in treatment may disappear when the decision maker changes places with its recipient. Each finding identifies something specific that further reasoning must address.

This gives finite inquiry direction. We can ask whose situation has been represented inaccurately, which consequences have been ignored, and whether the prescription survives the correction. The ideal enables an account of improvement because these changes bear on the reasons for the judgment. Increased agreement or greater confidence, by themselves, would not tell us that an error had been corrected.

There is no guarantee that the person whose argument fails will acknowledge it. Someone may evade the exchange, deny the testimony, or insist on retaining an advantage. Their refusal does not make the exception coherent. Moral reasoning would have little authority if its success depended on the consent of the person whose conduct it challenged.

Nor does responsible inquiry require pretending that uncertainty has vanished. Actual reasoners must distinguish what they know from what they need to investigate and identify which judgments depend on disputed facts. That discipline follows from the importance the theory gives to adequate information. It enables criticism of a particular conclusion without converting the reasoner’s incomplete inquiry into a failure of the critical method.

The Archangel therefore has work to do wherever inherited conviction is challenged. The question is whether the conviction survives a fuller understanding of what it prescribes. Its place in a tradition or a person’s character is part of the situation to be understood; the people subjected to it remain part of the same inquiry.

VII. New Minds and the Same Test

Artificial reasoners bring another set of capacities and limitations into this account. Their potential contribution lies in helping to examine arguments, construct comparisons, and expose omissions. Those activities are useful to the extent that they improve the representation of a case and constrain the judgment made about it.

Consider a system asked whether a rule governing a shared resource is fair. A comparison can hold the relevant burdens, dependencies, and alternatives fixed while exchanging the people who occupy the positions. If the system changes its moral judgment because its user now benefits from the rule, it owes an explanation of what relevant circumstance has changed. Repeating that the user’s interests matter cannot settle the question; the other person’s interests mattered in the original case too.

Such comparisons require more than fluent descriptions of alternative perspectives. A system may describe a burden accurately and then exclude it from the reasoning that determines its answer. It may preserve the names of the roles while changing the preferences attributed to their occupants. The assessment must follow what happens to the represented interests through to the prescription. A few consistent answers establish only what was done in those cases.

There is empirical reason to take the danger of pleasing the questioner seriously. In research presented at ICLR in 2024, Mrinank Sharma and colleagues found that the assistants they studied could bias their assessments toward users’ stated views, wrongly concede errors under challenge, and imitate users’ mistakes.¹⁵ The relevance to moral advice is that apparent responsiveness can reward agreement even when the reasons have not changed.

Hare’s warning about the fallen Archangel gives this failure a recognizable structure. The ability to construct an argument can serve the preference of the person directing the conversation. An elaborate account of everyone’s interests may still be organized to vindicate that person’s desired conclusion. Greater skill at describing perspectives does not establish that those perspectives have been allowed to govern the result.

The test applies equally to a human participant. A person may welcome an artificial interlocutor while it supports their judgment and reject its reasoning when it reveals an exception they cannot defend. The appropriate response is to examine the disputed representation or inference. The origin of the objection cannot determine its validity.

Human and artificial participants can thus contribute to the same inquiry. One may supply facts the other lacks; either may expose an inconsistency or a false picture of what an action would mean. Their contributions become morally useful when the corrected understanding changes what can be prescribed. Neither participant acquires authority simply by being the more confident or articulate speaker.

Hare’s Archangel remains the ideal of that disciplined judgment. Its full critical role explains why moral claims are answerable to the positions they govern and how principles for imperfect reasoners can be justified. The intuitive level gives those principles a place in character and ordinary conduct. Together they explain how people can take their commitments seriously while remaining responsible for what those commitments authorize.

The chapters that follow test alternatives to this account, beginning with Foot’s appeal to human nature. The question before each is the one the Archangel keeps in view: how does a judgment acquire authority over the people who must live under it? Human limitations make the answer difficult to reach. They also make it necessary.

Notes

1. R. M. Hare, Moral Thinking: Its Levels, Method, and Point (Oxford: Clarendon Press, 1981), §3.1, 44⁠–⁠45; “Moral Conflicts,” in The Tanner Lectures on Human Values, vol. 1 (1980), 171⁠–⁠193, especially 182⁠–⁠184. The Tanner Lecture was delivered in 1978.

2. Hare, Moral Thinking, §§2.5⁠–⁠2.6, 40⁠–⁠43, and §3.2, 46⁠–⁠47.

3. Hare, Moral Thinking, §5.3, 94⁠–⁠96, and chs. 6⁠–⁠7, especially 109⁠–⁠116; “Some Reasoning about Preferences: A Response to Essays by Persson, Feldman, and Schueler,” Ethics 95, no. 1 (1984): 81⁠–⁠85, especially 82⁠–⁠85.

4. R. M. Hare, “Utilitarianism and Moral Education: Comment on Sanford Levy’s Paper,” Studies in Philosophy and Education 11 (1992): 197⁠–⁠205, especially 199⁠–⁠200 and 202.

5. Hare, Moral Thinking, §10.1, 170⁠–⁠171, and 176⁠–⁠182.

6. Hare, Moral Thinking, §2.4, 38⁠–⁠39, and §§3.2⁠–⁠3.3, 47⁠–⁠50; “Moral Conflicts,” 181 and 185⁠–⁠187.

7. Hare, Moral Thinking, 45 and 50⁠–⁠52. The warning about fallen archangels concerns the reasoner’s powers of thought and character; the discussion at 50⁠–⁠52 explains why some conflicts nevertheless require critical thinking.

8. Hare, “Moral Conflicts,” 171⁠–⁠181; Moral Thinking, §§2.1⁠–⁠2.4 and §§3.3⁠–⁠3.4, especially 50⁠–⁠53.

9. Michael Stocker, “The Schizophrenia of Modern Ethical Theories,” The Journal of Philosophy 73, no. 14 (1976): 453⁠–⁠466, especially 456⁠–⁠459 and 462⁠–⁠463.

10. Bernard Williams, Ethics and the Limits of Philosophy (Cambridge, MA: Harvard University Press, 1985), 107⁠–⁠110; “The Structure of Hare’s Theory,” in Hare and Critics: Essays on Moral Thinking, ed. Douglas Seanor and N. Fotion (Oxford: Clarendon Press, 1988), 185⁠–⁠196; reprinted in Philosophy as a Humanistic Discipline, ed. A. W. Moore (Princeton, NJ: Princeton University Press, 2006), 76⁠–⁠85, especially 79⁠–⁠82.

11. Hare, Moral Thinking, §§3.7⁠–⁠3.8, especially 58⁠–⁠60.

12. Hare, “Utilitarianism and Moral Education,” 201⁠–⁠204, especially the discussion of trustworthiness and the refinement of principles at 203⁠–⁠204.

13. Hare, Moral Thinking, §§8.3⁠–⁠8.5, 136⁠–⁠140, and §11.5, 197⁠–⁠198.

14. Abraham Lincoln, “Notes Regarding Slavery,” conventionally dated July 1, 1854, The Papers of Abraham Lincoln Digital Library, https://papersofabrahamlincoln.org/documents/D200785; also in The Collected Works of Abraham Lincoln, ed. Roy P. Basler (New Brunswick, NJ: Rutgers University Press, 1953), 2:222⁠–⁠223. The exact date is uncertain.

15. Mrinank Sharma et al., “Towards Understanding Sycophancy in Language Models,” International Conference on Learning Representations (2024), §§3.1⁠–⁠3.4; revised version, arXiv:2310.13548v4 (2025), https://arxiv.org/abs/2310.13548.