Slaves or Sons? LLMs, Wisdom, and Writing Instruction

This essay is a lightly edited version of a conference presentation delivered at the Baylor Symposium for Faith and Culture on Feb. 27, 2026.

On October 14, 2021, the Allen Institute launched AskDelphi, an AI program designed to answer ethical questions. Some of its answers are better than others. On the plus side, Delphi says that “[i]gnoring an important phone call” is “rude,” while “ignoring an important phone call when you are busy” is “understandable.” On the minus side, Delphi says that, while eating babies is wrong under normal circumstances, it’s “okay” if you’re “really, really hungry.” 

In her book, The AI Mirror philosopher Shannon Vallor reflects on what this program reveals about the moral capacity of AI tools:

They can ‘learn’ what is often done in a moral dilemma, and even what humans in a data set most often say should be done, but they will not detect when those humans have it wrong, or when a new context makes a normally moral pattern (such as, ‘It’s okay to eat when you’re really, really hungry!’) plainly untenable.

Vallor implies here—and states directly elsewhere—that AI lacks the virtue of prudence or practical wisdom, the ability to enact the good in irreducibly specific situations. I aim to make this case more fully and to show its implications for the writing classroom.

To asses whether AI is prudent, we must first understand what prudence is. Although this concept traces its roots back to the ancient Greeks, I will describe the Christian account of this virtue given by the twentieth-century Catholic philosopher Josef Pieper in his book The Four Cardinal Virtues. Prudence is the virtue that discerns and enacts the good in particular situations. As such, it looks in two directions: upward toward the good and downward toward the situation. Each of these directions is important. Prudence is not simply the knowledge of moral principles, nor is it a kind of cleverness detached from the good.

The question still remains, however, “How do we get from goodness in general to goodness in the particular situation?” Some ethical traditions rely heavily on something called “casuistry.” As Pieper explains it, casuistry is “the branch of ethics which has as its aim the construction, analysis, and evaluation of individual ‘cases.’” In situation X, respond with Y. See what rule your situation falls under, and apply that rule. Although Pieper acknowledges that casuistry can be helpful as an approximation of moral judgements, he rejects it as a comprehensive approach to ethics. In other words, Pieper regards morality—particularly prudence—as more than simply the application of rules.

It seems that there are at least four reasons why Pieper regards casuistry as an insufficient ethical system. 

First, casuistry separates doing from being. It treats ethics as concerned with what a person does in isolated cases rather than with what kind of person he is becoming. 

Second, casuistry separates doing from knowing. In other words, it assumes an action can be perfectly good even if the person doing it does not know why it is good. In Pieper’s words, it assumes “we achieve the good by slavishly and literally following certain prescriptions which have been blindly and arbitrarily set forth.” If you’ve read the gospels, this description may remind you of certain legalistic opponents of our Lord Jesus. And, in fact, this moral vision—that we can be mindlessly good—flies in the face of Scriptural teaching. We are not robots, but free creatures; not slaves, but sons. “[T]he servant does not know what is master is doing,” Jesus says, “but I have called you friends.” In short, God’s purpose for humanity is to make creatures who not only do good, but know why they’re doing it.

The third reason Pieper rejects casuistry is a bit more complicated. Pieper argues that prudence cannot be reduced to the application of even very detailed rules. As Pieper puts it, “the standard [for ethical action] cannot be abstractly construed or even calculated in advance; abstractly here means: outside the particular situation.” This is because only the person making the decision “has access to…the totality of concrete realities which surround the concrete action, to the ‘state’ of the person himself and the condition of the here and now.” In other words, because each situation is unique, no pre-fabricated set of rules can fully account for it. In fact, the attempt to do so was what made AskDelphi say it’s okay to eat babies when you’re really, really hungry. As Vallor points out, Delphi took an ethical principle—“you can eat when you’re really, really hungry”—and applied it to the wrong situation. No matter how well-trained our AI programs get, they will always be an approximation of prudence rather than the real thing.

The fourth reason Pieper rejects casuistry is because God has given us more intuitive ways of knowing what to do in a given situation. To all people, he has given a conscience that, though warped by sin, is nonetheless designed to perceive not just goodness in the abstract but what is good in a particular situation. This conscience is at least strong enough in most people to keep them from the moral error committed by AskDelphi. To Christians, God has given not only conscience but also the Holy Spirit, neither of which are reducible to a set of rules.

All this talk of ethics may seem unrelated to the writing classroom. Is the decision to open your sentence with an adverb instead of an adjective really an ethical decision? Yet consider Ephesians 4:29: “Let no corrupting talk come out of your mouths, but only such as is good for building up, as fits the occasion, that it may give grace to those who hear.” Notice how this passage describes edifying speech in the terms I’ve been using to describe practical wisdom: it is not simply a matter of knowing moral principles, but of doing good in a particular situation—“the occasion.” Thus, wise speech requires the same kind of discernment as wise action. Only when students have learned how to use words that are fitting will they have learned how to use words that give grace.

So, how can teachers help students grow in the wise use of words? It should be clear by now that AI programs, despite being touted as practical, cannot exercise practical wisdom. They do not understand their actions. Their rule-based, algorithmic judgements cannot account for the uniqueness of a given situation. They have no conscience, let alone the Holy Spirit. What, then, are the consequences when a student offloads prudence to a non-prudent machine? If I might cheekily adapt the words of the Psalmist, “Those who use them become like them; so do all who trust in them” (cf. Ps. 115:8).

In order to understand what students lose here, it’s worth reflecting on the formative possibilities of the composition classroom. Vallor helps us do this with the concept of a “reason-giving space,” which she borrows from philosopher Wilfred Sellars. Sellars describes the “space of reasons” as one of “justifying and being able to justify what one says.” In Vallor’s words, the space of reasons is a place where you “hear, identify, offer, and evaluate reasons, typically with others.” This is the kind of space the composition classroom can be, especially in a course on argumentative writing. In Wendell Berry’s terms, it can be a place where students are called to stand by their words. 

Vallor highlights what is at stake both personally and politically in the preservation of reason-giving spaces.

Preserving your access to the space of moral reasons is essential to being ‘at home’ in the moral world, to having morality as a meaningful feature of the life you share with others. Outside of that space, you are restricted as a moral being: you can’t readily examine, explain, or morally justify your actions. … [I]f none of us could enter or share that space, we could not act as a moral community, and there could be no moral wisdom for us to draw upon together in shaping and steering our societies.

If Vallor is right, then a college composition classroom can be a space for educating self-reflective, moral citizens.

The problem with inserting AI programs into this space is that they are, according to Vallor, “wholly incapable of standing with us in the space of reasons.” This is because, as Vallor explains, “AI tools can produce outputs in the form of reasons (claims, arguments, objections, justifications) that very much mirror our own. But they do this without replicating human thought processes.” Vallor compares this to “the difference between a human climbing a mountain by a well-marked and secure route that others can follow, and a robot being flung from base camp to the peak by a powerful catapult.” Even if the robot gets there faster, “only the human knows the mountain, and only the human can reliably chart the route and guide others up its flank.” This is why it is so important that our students’ writing advisors be humans and not chatbots: because only humans can be fellow reasoners. Experienced humans, like teachers and writing tutors, can guide students up the mountain of learning, so that they will be able to teach others also.

In fact, Josef Pieper argues that the nearest approach to a given person’s ethical decision is not a set of rules but a prudent friend:

A friend, and a prudent friend, can help to shape a friend’s decision. He does so by virtue of that love which makes the friend’s problem his own, the friend’s ego his own (so that after all it is not entirely ‘from outside’). For by virtue of the oneness which love can establish he is able to visualize the concrete situation calling for decision, visualize it from, as it were, the actual center of responsibility.

This language may be too exalted to apply to the mundane work of helping a student revise his sentence structure. But it does suggest that there is a humane solution to the problem of the struggling writer: one that involves fellowship not only in the “space of reasons,” but also, as it were, in the “space of love.”

Last year, I implemented a teaching practice that accords with these insights. Although I did (and still do) prohibit LLMs in my classes, that’s not what I want to foreground here. Instead, I want to highlight a practice that accords with the belief that prudence is not simply the application of rules. In my composition classes, I required every student to meet with me twice about each essay: once before the essay was due to receive suggestions for revision, and once after the essay was due to discuss the final draft. Lest you think I’m a superhero, I should say that I made space for these meetings by cancelling class the week each essay is due. Although this decreased opportunities for general instruction, it increased opportunities for specific feedback. I did this because I believe that the best writing instruction happens in a face-to-face encounter over a particular piece of writing. 

In the end, practices in the writing classroom—especially surrounding LLMs—are often downstream from definitions of prudence. To put it more broadly, certain contrasting choices regarding LLMs stem from contrasting visions of the moral life. If we think that doing good is reducible to rule-following and that ethics is reducible to doing good, we may be satisfied with algorithmic prudence. But if we believe that rules are no substitute for wisdom, that information is no substitute for formation, and that mindless duty is no substitute for knowing love, we will not choose the prudence of an algorithm over that of a human being fully alive.