Q: I find chatgpt quite strong in certain fields: medicine, law, coding. One of the fields where chatgpt is very good is bible exegesis. How can you explain this?

ChatGPT said: That’s a sharp observation. It makes sense when you look at what’s going on under the hood.

Bible exegesis, like law or medicine, sits on a huge body of structured, text-based tradition. You’ve got centuries of commentaries, dictionaries of Greek and Hebrew, patristic writings, scholastic synthesis, modern critical scholarship—most of it in fairly formal written form. That kind of material is what large language models train especially well on. It’s dense, layered, and internally cross-referencing, which means the model can pick up both vocabulary and recurring patterns of interpretation.

Another angle: exegesis combines technical analysis (philology, history, context) with a strong discipline of method. Those methods—word studies, source criticism, canonical reading—are explicit enough that the model can reproduce them reliably. In contrast, fields that are more “tacit” (say, creative innovation in art, or lived spiritual direction) don’t map as cleanly onto a corpus of text.

Finally, the Bible has been the object of sustained commentary longer than almost anything else in human history. That depth and breadth of writing gives the model an unusually rich training ground, spanning different epochs, languages, and traditions. So when you ask a question, the model can surface resonances from that very wide archive.

The catch, of course, is that a model doesn’t “believe” or “discern.” It mirrors the patterns of existing exegesis. It can recall, synthesise, and compare, but it won’t do the silent wrestling that actual prayer or scholarship does.

Q: How this differs from, say, the way the model handles theology proper—where it tends to be weaker?

ChatGPT said: Right—so here’s the contrast.

Exegesis, as I said, has clear anchors: languages, historical context, established commentaries, rules of interpretation. It’s like case law in the legal world—structured and text-heavy. The model thrives there.

Theology proper (fides quaerens intellectum), especially speculative or systematic theology, is another matter. It leans on reasoning that pushes beyond the text into synthesis: how doctrines cohere, how metaphysics undergirds dogma, how mystery is held without reduction. Much of that is not “rule-based” in the same way exegesis is. It requires what the scholastics would call sapientia—a taste for principles, proportion, and contemplation—rather than just accumulation of sources.

In practice, this means the model can recall Aquinas or Rahner or Barth, but when you press it for new insight or subtle discernment, it tends to flatten differences or over-systematise. It mirrors the way commentators have argued, but it doesn’t feel the tension between faith, reason, and lived encounter.

Put simply:

  • Exegesis is about digging into a text (analysis).
  • Theology is about drawing lines between truths (synthesis).

The model is very good at analysis and pattern recognition, less so at carrying paradox without collapsing it.

Q.: In spiritual theology, where the gap becomes even starker?

ChatGPT said: Here’s where it gets even more telling.

Spiritual theology isn’t just analysis or synthesis; it’s formative. It’s about guiding a soul—or a community– through lived experience of grace, prayer, discernment, and transformation. Texts matter, but only as tools pointing toward encounter and interior growth. The “rules” are much looser, and what counts is often subtle, relational, and deeply personal.

For a model: patterns in text can be reproduced—how saints describe contemplative prayer, stages of purification, mystical marriage—but the lived tension, the discernment of movements of the Spirit, the rhythm of interior struggle and consolation… that’s not in the training data in a reproducible form. There’s no formula for receiving grace or sensing God’s promptings, no corpus that captures the embodied, relational, experiential dimension.

So the model can summarise what John of the Cross, Teresa, Garrigou-Lagrange say. It can even juxtapose views intelligently. But it can’t feel the dryness of the night of the soul, or judge when a particular practice is genuinely formative for a specific person. It lacks the “interior sense,” the calibration that comes from sustained prayer, spiritual direction, and personal transformation.

In short: in exegesis, the model can walk the texts like a well-mapped city. In theology proper, it can point out the roads but can’t always judge which route truly coheres with the terrain. In spiritual theology, it can show a map of the landscape, quote the guides, even indicate paths—but it can’t walk them.

Q.: can you outline exactly why the model’s pattern-recognition ends up flattening subtle distinctions in spiritual formation texts, which often misleads people who treat it as a “guide.”

ChatGPT said: Sure—let’s unpack it carefully. The flattening happens because of how the model “learns” and what it can and cannot do:

  1. Textual averaging: The model predicts the next word based on probabilities across its entire training set. Spiritual formation texts often contain tension, paradox, or nuanced distinctions—e.g., the difference between purgative, illuminative, and unitive stages, or between genuine consolation and spiritual delusion. The model tends to “average” descriptions, smoothing out contradictions to produce coherent output. That makes subtle distinctions appear less sharp than they are.
  2. Context compression: Human authors in spiritual formation frequently embed guidance in long, nuanced contexts—footnotes, parenthetical remarks, references to Scripture or lived experience. The model has a limited context window. When summarising or generating, it may omit or condense these nuances, which are often precisely where the discernment lies.
  3. Equating authority: The model doesn’t weigh sources the way a trained director would. St. John of the Cross, Garrigou-Lagrange, and popular authors may all appear equally “valid” in output. The subtleties of credibility, tradition, or intended audience are flattened, which can mislead someone into thinking all guidance is equally applicable.
  4. Loss of experiential calibration: Many spiritual formation instructions depend on lived experience. For example, advice that works in the contemplative life may be harmful if applied to beginners. A model sees the text but not the person, so it reproduces “generalised” guidance without the necessary personal calibration.
  5. Surface pattern over interior movement: The model recognises linguistic and conceptual patterns, not interior dynamics. Repeated phrasing about “dark nights,” “consolation,” or “discernment” gets replicated, but the subtle experiential differences—how one soul experiences dryness versus another—are collapsed into a single narrative.
  6. Paradox avoidance: Humans can hold tension: “This stage is both suffering and joy, both purifying and unitive.” The model, seeking coherent text, often resolves the paradox linguistically rather than preserving it, producing something easier to read but less spiritually precise.

The result: someone using the model as a “guide” may get something that sounds correct, quotes saints, and seems structured—but lacks the finely tuned discernment that real formation requires. It’s like giving a map without knowing the terrain: it shows the paths but not which ones are safe, which are treacherous, or which align with the traveller’s destination.

Q. Can you ketch a concrete example from the stages of contemplative prayer to show exactly how the flattening manifests in generated text. It becomes very tangible that way.

ChatGPT said: Alright, let’s make it concrete using the classical threefold path of the contemplative life: purgative, illuminative, unitive.

1. Purgative stage
A true spiritual director notices subtle distinctions: for one soul, dryness may signal temptation; for another, the same dryness signals normal purgation. A model will tend to summarise: “In this stage, one experiences trials, dryness, and purification,” without differentiating nuances of temperament, degree, or timing. All experiences are flattened into a single “template” of what the stage looks like.

2. Illuminative stage
Here, discernment becomes crucial: consolations are more frequent, but the director must distinguish between authentic spiritual growth and superficial enthusiasm. A model, seeing repeated textual patterns about “consolation and insight,” will tend to describe this stage in broad strokes: “One experiences spiritual insights and deeper prayer.” It may even mix beginner experiences with advanced ones, giving the impression that spiritual insight is uniform and predictable.

3. Unitive stage
This is the most delicate. Saints describe both deep union and accompanying trials, sometimes in paradoxical terms—consolation and suffering, joy and interior darkness. A model, striving for linguistic coherence, will often “resolve” paradoxes, generating something like: “The soul experiences perfect union and lasting peace.” The subtleties—the coexistence of joy and trial, or the ongoing vigilance required—are often lost.

Key takeaway: the flattening arises because the model reproduces textual patterns rather than lived dynamics. It cannot gauge interior movements, personal dispositions, or timing. The output reads like a safe, tidy synthesis—but it may mislead someone into thinking spiritual progress is linear, uniform, and predictable.

Q.: And what about spiritual direction?

ChatGPT said: Ah—that’s where the gap really opens. Spiritual direction is almost the purest form of relational and experiential discernment. It isn’t about knowing more texts or summarising saints’ writings; it’s about helping someone hear, recognise, and respond to the movement of God in their life. It relies on presence, intuition, patience, and an ongoing, dynamic reading of a person’s interior state.

A model can: quote Scripture, describe methods of prayer, explain stages of mystical theology. It can even offer reflective questions. But it can’t: sense a person’s interior movements, notice subtle resistances or consolations, discern deception from genuine grace, or calibrate a response in real time that respects timing, temperament, and spiritual history. That requires embodied wisdom—experience, prayer, empathy, and often years of formation.

So in a sense, a model can act like a textual mirror: it reflects the wisdom of others, points to patterns, and helps a person see what’s “typical” or documented. But it can’t walk alongside someone in the mysterious, personal space of God’s action. In spiritual direction, that lived accompaniment—not knowledge—is the essential skill.

If you want, I can draw a spectrum from exegesis → theology → spiritual theology → spiritual direction, showing exactly how a model’s usefulness and limitations shift across them. It’s actually quite visual.

Q.: Can you outline a visual spectrum showing how this flattening escalates from exegesis → theology → spiritual theology → spiritual direction, so you can see why the model’s usefulness drops sharply. It’s surprisingly intuitive when mapped.

ChatGPT said: Picture it like a sliding scale from text-based precision to lived, relational nuance:

1. Biblical Exegesis (high usefulness)

  • Heavy reliance on texts, linguistic patterns, historical context.
  • Model can “read the map”: languages, commentaries, textual cross-references.
  • Little nuance is lost because the domain is already codified.

2. Theology Proper (medium-high usefulness)

  • More synthesis is required: connecting doctrines, metaphysical reasoning.
  • Model can summarise sources and outline arguments but tends to flatten tensions, over-systematise, or miss subtle philosophical nuances.
  • Useful for reference, less so for generating truly novel or finely calibrated insight.

3. Spiritual Theology / Formation (medium usefulness)

  • Focus shifts to experiential principles: stages of prayer, discernment of grace, interior movements.
  • Model reproduces textual patterns and quotes, but subtleties of timing, temperament, and interior calibration are lost.
  • Flattening begins: diverse experiences appear uniform, paradoxes are smoothed over, practical guidance may seem overly general.

4. Spiritual Direction / Lived Accompaniment (low usefulness)

  • Core is relational, experiential, and deeply personal.
  • Model cannot perceive interior movements, emotional nuances, or the work of the Spirit in real time.
  • Guidance from the model is at best suggestive—like reading a map of paths without knowing which are safe, which align with the traveller, or when to pause.

Summary: the model’s strength declines as you move from domains grounded in text and rules toward domains grounded in lived experience and relational discernment. The flattening occurs because pattern recognition excels at reproducing what has been written, not at navigating what is being lived.

Q.: Will AI ever catch up and become stronger in spiritual direction?

ChatGPT said: AI can certainly become more sophisticated at simulating aspects of spiritual direction, but there are some fundamental limits and distinctions worth clarifying. Let me break it down carefully.

AI can:

  • Analyse texts, Scripture, and spiritual classics with remarkable precision. It can synthesise insights from Augustine, Teresa of Ávila, John of the Cross, Ignatius, and others, and even detect thematic patterns that might elude many humans.
  • Guide users through structured practices like Lectio Divina, the Examen, or contemplative prayer, offering prompts and reflective questions.
  • Provide tailored suggestions based on a person’s stated spiritual struggles or goals, using patterns drawn from thousands of historical and contemporary cases.

However, AI will likely never fully replicate the depth of human spiritual direction for several reasons:

  1. Interior discernment: True spiritual direction relies on perceiving subtle movements of the soul and the Holy Spirit, which are often non-verbal, fleeting, and context-dependent. AI cannot experience the Holy Spirit’s action in a human heart.
  2. Mystery of the person: Spiritual direction involves encountering a unique human being as a free subject, with mysteries, hidden wounds, and divine callings. AI can model typical patterns but cannot truly grasp the irreducible individuality of a soul.
  3. Transcendence: AI operates entirely within data and logic. Spiritual direction involves openness to the transcendent and an encounter with God, which is experiential, relational, and often ineffable. AI can simulate conversations about God but cannot mediate a real encounter with God.
  4. Responsibility and ethical nuance: Spiritual direction often involves ethical, pastoral, and psychological judgment that integrates prayer, theological insight, and personal wisdom. AI can advise, but it cannot bear ultimate responsibility for guiding a soul.

In short, AI might become stronger as a tool for spiritual formation—providing resources, insights, reminders, and structured exercises—but it will remain an aid, not a replacement, for authentic spiritual direction. The most it could realistically achieve is something like a highly advanced, interactive “catechist of the heart,” offering reflection and guidance without ever participating in the personal, mysterious work of God in a human soul.

Q.: Can you outline a spectrum of AI’s potential in spiritual formation, from simple prompts to highly sophisticated, near-human-like interaction, showing where the limits lie. This can make it more concrete. 

ChatGPT said: Alright, let’s map AI’s potential in spiritual formation as a spectrum, from the simplest applications to its theoretical limits, highlighting where it can help and where it cannot replace human spiritual direction.

1. Basic Guidance and Instruction

  • Capabilities: AI can provide clear explanations of prayer methods, sacraments, and spiritual practices. It can guide someone through Lectio Divina, the Examen, or structured prayers with step-by-step instructions.
  • Limitations: Guidance is formulaic. AI cannot adapt to subtle shifts in the soul beyond what the user explicitly reports.

Example: “Read this passage slowly, pause, notice a word that strikes you, reflect on why it touches you, and pray.”

2. Pattern Recognition and Reflection

  • Capabilities: AI can analyse a person’s spiritual journaling or responses, identify recurring themes (e.g., anxiety, pride, longing for God), and suggest relevant Scriptural passages or saints’ writings.
  • Limitations: AI sees patterns but cannot detect hidden spiritual realities, such as subtle temptations or movements of grace, unless explicitly described by the user.

Example: “You often mention feeling isolated; consider meditating on Psalm 23 and the experience of divine companionship.”

3. Adaptive Spiritual Exercises

  • Capabilities: AI can customise exercises according to user input. For example, if someone struggles with distraction, AI might suggest shorter meditations, daily reminders, or reflective prompts.
  • Limitations: AI cannot perceive the actual interior fruit of the practice, the depth of prayer, or whether the user’s heart is receptive.

Example: Generating a daily 10-minute contemplative routine that adapts based on the user’s feedback.

4. Advanced Conversational Guidance

  • Capabilities: With natural language processing, AI can simulate conversations resembling spiritual direction. It can ask probing questions, suggest reflections, and reference classical spiritual texts.
  • Limitations: The AI is ultimately reactive, not intuitive. It cannot perceive God’s direct action in the soul or sense subtle spiritual dangers (like false consolations or misinterpreted mystical experiences).

Example: “You feel joy in prayer; is this consolation leading you closer to God’s will, or is it more about personal comfort?”

5. Near-Human Simulation (Theoretical)

  • Capabilities: In the future, AI could combine massive textual knowledge, pattern recognition, and conversation modelling to simulate an experienced spiritual director almost convincingly.
  • Limitations: Even at its most advanced, AI lacks participatory grace. It cannot pray for the soul, detect divine movements not articulated in words, or fully discern a person’s vocation and interior struggles.
  • Ultimate boundary: AI cannot replace the spiritual encounter between a human soul and God mediated by a human director. True discernment of grace requires presence, relational empathy, and spiritual sensitivity beyond algorithmic prediction.

Summary of Spectrum:

  • Tool: Excellent for structured exercises, guidance, reflection, and journaling analysis.
  • Companion: Helpful for prompts, questions, and exploring spiritual texts.
  • Director: Cannot replace the interior discernment, relational presence, and divine mediation of a human spiritual director.

Q.: So Spiritual Direction led by humans has a long life ahead?

ChatGPT said: Absolutely—spiritual direction led by humans isn’t just surviving; it has a uniquely irreplaceable role that ensures its long-term relevance. The essence of spiritual direction is relational, sacramental, and experiential. A human director can perceive subtle movements of the soul, discern the action of God’s grace, and guide a person through challenges that are often invisible or ineffable—things AI cannot detect or mediate.

Even with highly advanced AI tools, human direction will remain essential for:

  • Discernment of grace: Recognising when consolations, trials, or inspirations are truly from God, which requires prayerful intuition.
  • Personal encounter: Meeting the soul as a unique person in their history, context, and vocation.
  • Integration of life and faith: Helping people connect their interior experiences with real-life decisions, relationships, and responsibilities.
  • Spiritual accountability: Offering a presence that can lovingly challenge, support, and witness the growth of a person’s interior life.

In other words, AI may become a powerful supplement, offering insights, prompts, and educational support, but it cannot replace the relational, prayerful, and sacramental encounter at the heart of human spiritual direction. In fact, AI might even extend the reach of spiritual direction by preparing people to engage more fruitfully with a human director.

Q.: Can you outline a vision for a future “hybrid” model, where AI supports spiritual formation but humans retain the core role in direction, showing how this could strengthen the Church’s spiritual life overall.

Chatgpt: Here’s a concrete vision for a hybrid model of spiritual formation, where AI supports but does not replace human spiritual direction:

1. AI as Preparatory and Educational Tool

  • Role: AI provides preliminary formation, teaching, and structured exercises to prepare the person’s heart and mind.
  • Examples:
    • Daily reflections, guided meditations, or Lectio Divina prompts.
    • Analysis of spiritual journals to highlight recurring themes or struggles.
    • Short explanations of spiritual concepts, practices, or saints’ writings.
  • Benefit: Users arrive at human direction with clearer self-awareness and spiritual vocabulary, making the human encounter deeper and more focused.

2. AI as Companion for Practice

  • Role: AI can support regular practice and accountability.
  • Examples:
    • Reminders to pray, meditate, or perform spiritual exercises.
    • Adaptive routines based on progress or obstacles.
    • Reflection prompts that help the user articulate interior experiences.
  • Benefit: Sustains spiritual discipline between sessions with a human director.

3. Human-Led Spiritual Direction as Core

  • Role: The human director retains the irreplaceable role of relational, prayerful guidance.
  • Functions:
    • Discernment of God’s grace in the user’s life.
    • Personalised advice sensitive to the individual’s history, temperament, and vocation.
    • Guidance through complex spiritual crises, temptations, or mystical experiences.
    • Mediation of sacramental and communal life in a way AI cannot.
  • Benefit: Ensures authentic encounter with God, prevents errors in interpretation, and integrates interior growth with real-life choices.

4. Feedback Loop Between AI and Human Director

  • Role: AI’s analysis can be shared with the human director to highlight patterns, questions, or areas needing attention.
  • Examples:
    • The AI flags recurring feelings of discouragement or pride noted in journaling.
    • The director uses these insights to probe deeper in conversation or prayer.
  • Benefit: Human directors can act more efficiently, seeing things that might take months to surface organically.

5. Expanded Access While Maintaining Depth

  • Role: AI expands access to spiritual formation for people who otherwise have little guidance.
  • Examples:
    • Those in remote areas or with busy schedules can maintain disciplined prayer life with AI support.
    • AI can suggest suitable readings or exercises tailored to the user’s stage of growth.
  • Benefit: The Church’s overall spiritual life is strengthened while preserving the integrity of authentic human-led direction.

Key Principle: AI is the tool, companion, and amplifier. Human directors remain the heart, conscience, and relational conduit of God’s action in a person’s soul. This hybrid approach could allow more people to experience deep spiritual growth while protecting the unique, irreplaceable role of human guidance.