Seminary Leaders Warn: AI Cannot Replace Spiritual Formation
Theological educators must treat artificial intelligence as a capable research tool while protecting the spiritual development that lies at the heart of seminary training, according to two leading voices in evangelical theological education.
Dr. Walker Tzeng, executive director of the World Evangelical Theological Institute Association, and Dr. John Dyer, dean of educational technology at Dallas Theological Seminary, presented their framework at an international webinar hosted by the International Council for Evangelical Theological Education. More than 70 educators from different regions attended.
The core issue separates what AI can do from what it fundamentally cannot. AI systems operate by statistically predicting which word follows the last one - a mechanical process that produces fluent text without understanding, lived experience, or relationship with God.
What AI Actually Does
Tzeng explained that large language models work through mathematical proximity. When a model encounters "king," it calculates that "queen" is statistically close and generates that word. The system synthesizes existing material rapidly and correlates ideas across sources.
What it cannot do matters more for seminary contexts: originate thought, draw on lived faith, or produce writing that emerges from a person's relationship with God, their church community, and Scripture.
"When we do write something, we draw from Scripture, or personal experience, we have our spiritual life or church community or peer review," Tzeng said. "We are also people that repent. We have faith and we have love and we draw from that as we do our theology."
AI has no access to any of that. Because AI cannot have a relationship with Jesus Christ, the content it generates remains an imitation of human theological writing rather than an expression of it.
Tzeng described AI as a capable research assistant for someone who already knows their subject - useful for processing and synthesizing existing material, less useful for generating genuine insight.
A Biblical Framework for Technology
Dyer approached the same questions through Scripture, tracing five chapters of the biblical narrative: creation, the image of God, the fall, the life of Jesus, and the new creation.
In Genesis, God gives humanity a dual mandate: to fill the earth and to cultivate and tend it. Innovation and preservation work together as responsible making. Tool use and creativity are not responses to sin - they are part of the original vocation.
On the image of God, theologians identify three categories: substantive (capacity to reason), functional (dominion over creation), and relational (unique bond with God and each other). AI can appear to encroach on the first two. It processes information faster and manages tasks more efficiently. The third category remains untouched.
"Our unique relationship with God is unique - AIs don't have that," Dyer said. "As good as they are at mimicking a lot of what humans do, they aren't really a 'they' in that sense."
He flagged a pastoral concern: people forming what feels like a deep relationship with an AI system while their relationships with other people and with God quietly fade.
The fall introduces a recurring temptation: using what we make as substitutes for God rather than expressions of faithful service. Dyer connected this directly to AI. "The temptation we face in some sense with AI right now is to have powers beyond our own ability - to be something more than human, and not to be God's image bearers, but to go beyond that."
Looking to Revelation's vision of the new creation - a holy city with roads, gates, and the accumulated work of human hands - Dyer argued that human making matters eternally to God. What people create is not discarded at the end of history but redeemed and reformed within it.
Practical Steps for Classrooms
Dyer recommended that educators first keep the purpose of seminary formation consistently in front of students, before any conversation about AI policy begins.
At Dallas Theological Seminary, he said, faculty tells prospective students that if they come primarily to acquire information, they can likely find it elsewhere for free. The real reason to come to seminary is to be transformed. "The goal is not to make the best chart of Leviticus possible," he said. "The goal is to become the kind of person who's made a chart of Leviticus - who's really had that become part of who they are."
He recommended distinguishing between different types of AI use rather than addressing the technology in broad strokes. Generating ideas, producing an outline, drafting text, and editing text are meaningfully different activities. Faculty should specify which are permitted and why.
Tzeng suggested reducing assignment word counts. Because AI excels at generating long, fluent essays, shorter prompts - 150 or 500 words rather than several thousand - force students to develop and compress their own thinking.
He also recommended raising grading standards on the assumption that students have AI available. If AI can produce text that reads at a doctoral level, grading students at their previous standard may no longer reflect genuine mastery.
Both speakers called for practical training in recognizing AI-generated errors, or "hallucinations." Tzeng said students who grasp the statistical nature of these systems are less likely to over-trust them.
The Ethics Question: Misrepresenting Capability
Dyer reframed the ethical issue beyond plagiarism toward a different kind of dishonesty: misrepresenting one's own capabilities.
"In the age of AI, probably what we're talking about more is presenting yourself as able to do something that you really can't do," he said. "If you're saying, 'I'm able to synthesize this idea' or 'I'm able to write this paper' when you really can't, that's where I think we're crossing an ethical boundary."
He applied the same logic to professors. Using AI to grade student work without disclosing it means educators are presenting themselves as doing something they are not.
DTS encountered this complexity when experimenting with AI voice generation as a substitute for subtitles in courses taught in multiple languages. A professor appeared to be speaking fluently in a language they do not know.
"If we're asking our students not to present themselves as able to do something they can't, but then we're having a professor being translated into another voice and literally presenting themselves as doing something they cannot do, that has some real conflict there," Dyer said.
Tzeng identified AI-generated citations as the most widespread integrity problem he has observed. Large language models cannot reliably produce accurate references. They fabricate sources, misattribute quotations, and cite real works incorrectly.
Where AI Helps and Where It Harms
Both speakers identified genuine benefits. Dyer described the shift from search-engine-based research to conversational AI as potentially productive, provided the tools link to verifiable sources.
DTS uses AI for practical automation: extracting data from PDFs into usable formats and drafting initial replies to high volumes of student emails during registration periods. A human reviews the AI-drafted response before it is sent.
"We're able to serve our students better," Dyer said. "Paying a little bit of money for an AI subscription versus an entire person in those really high moments - we're actually reducing the cost of education as a whole."
Tzeng noted that AI has been particularly useful for international students writing in English as a second language, helping them produce cleaner prose. But a corresponding harm exists: many of those students now spend more time, not less, working on their writing as they refine AI-generated text rather than developing their own voice.
The result becomes visible when those students preach. "If you are giving a sermon, you should really have your own voice in it," Tzeng said. "But someone's delivering a sermon and it just sounds like AI as they're delivering it - it's not very good."
Dyer framed beneficial and harmful uses on a spectrum running from full automation on one end to full preservation on the other. Tasks that are purely administrative or mechanical can appropriately be automated. Tasks that form the person doing them should be protected.
Theological education, as a discipline, belongs largely on the preservation side, particularly while students develop foundational skills. "When we move over into automation, that's where we start to lose a skill because we're giving that over to a machine to do," Dyer said.
The Decade Ahead
Dyer maintained that the core disciplines of theological training - biblical literacy, synthesis, discernment, pastoral judgment - matter most and are least amenable to automation.
"When you're facing someone in a hospital and they're asking you a difficult question, you can't turn to GPT in that moment," he said. "The goal of our instruction is love. It is not papers."
The challenge is keeping the main purpose of formation in front of students while acknowledging that once they are in ministry, AI tools may genuinely help them work more efficiently.
Tzeng suggested that emerging "agentic AI" systems capable of completing multi-step tasks autonomously could eventually assist with grading, one of the most time-consuming aspects of faculty work. Such a system could evaluate a student paper multiple times using different parameters and average the results.
He offered reassurance for theological educators concerned about the pace of change: theology and philosophy are among the disciplines least exposed to AI disruption compared to fields like mathematics or computer programming. The reason connects to the nature of theological knowledge itself.
"We gain our knowledge from God," Tzeng said. "We don't gain knowledge just from reiterating what everyone else is doing."
The Bottom Line
Clarity about what AI actually is must come before any decision about how to use it, Tzeng said.
"A lot of people interact with AI as if it's a human," he said. "But it's not - it's a tool. And we, as people made in God's image, have the responsibility to use it well."
The webinar is part of ongoing work within the International Council for Evangelical Theological Education to equip theological educators for a post-digital world. A recording of the session will be made available for educators who could not attend.
Your membership also unlocks:
