AI Must Be Judged as a Gospel Question, Not as a Morally Neutral Tool

Source: Darrell B. Harrison. "EP # 135 | AI and the Gospel - Just Thinking Podcast." September 8, 2025. justthinking.me

The Gist

Harrison and Walker argue that AI is a gospel issue because sinners make it and sinners use it. The first question is not “is it useful?” but “by what standard is it good?” Scripture, not cultural fashion, has to answer that. Don’t panic, and don’t treat it as a morally blank gadget.

Conclusion

Artificial intelligence is not a morally neutral instrument that Christians may evaluate only by usefulness. Because every technical system is designed, trained, and deployed by fallen human beings, AI is a theological and gospel matter: it raises the question of the standard of right and wrong, and it must be assessed by Scripture’s account of human nature, the image of God, wisdom, and sin. The fitting Christian posture is discernment rather than panic or uncritical embrace.

Premises

1. Normative questions about how a technology may be used always lead back to a meta-ethical question: what makes something right or wrong in the first place. Absent a transcendent standard, the answer collapses into whatever moral framework presently dominates the culture.
2. Scripture teaches that every human being is a sinner (Ecclesiastes 7:20; Romans 3:23). Any invention that bears the fingerprints of sinners can therefore be used for good or for grave evil. Solomon’s observation that God made humankind upright, but they have sought out many devices (Ecclesiastes 7:29), names this pattern: after the fall, human beings do not merely sin; they innovate their rebellion.
3. AI is not merely a hammer. It is designed to mimic human reasoning, decision-making, and creativity. If total depravity reaches human thought, then systems built to replicate that thought at superhuman scale will carry the biases, assumptions, and moral frameworks of unregenerate minds unless those systems are judged by God’s revealed will.
4. God’s word is sufficient to evaluate the technology even when it does not name every application. Scripture supplies the relevant principles: human nature, the image of God, the ordering of society, wisdom and folly. Fear is not required; discernment is.

Assumptions

• The Christian scriptures are a public, objective moral authority, not one private preference among others.
• “Not morally neutral” does not mean “intrinsically evil as an artifact.” It means the tool cannot be quarantined from the hearts and worldviews of its makers and users.
• Christians may still use powerful tools; the claim is about the standard of evaluation, not a ban.
• Source: Just Thinking Podcast, Ep. 135, “A.I. and the Gospel,” Darrell Harrison and Virgil Walker, published 8 September 2025. https://podcasts.apple.com/us/podcast/ep-135-a-i-and-the-gospel/id1328733796?i=1000725471889 Official episode page: https://justthinking.me/ep-135-ai-and-the-gospel/. This reconstruction steelmans the hosts’ claim from approximately 11:58–22:26 of the episode; it is not a verbatim transcript.

Analysis

Overall strength: Moderate. Argument type: Deductive.

Premise Strength

• P1 (meta-ethical collapse without transcendent standard) (Weak) — Internally serviceable given A1, but its exhaustive dichotomy between transcendence and cultural relativism is contested by well-established secular ethical traditions, making it the argument's weakest structural link and the point most vulnerable to rebuttal.
• P2 (universal sinfulness and the pattern of 'many devices') (Strong) — Given the granted assumption that Scripture is authoritative, this premise is well-supported by direct citation and is doctrinally uncontroversial within the Christian tradition; it functions largely as an application of A1 rather than an independently risky claim.
• P3 (AI mimics reasoning and inherits unregenerate biases at scale) (Moderate) — The underlying empirical observation — that AI systems reflect the assumptions embedded in their training data and design choices — is well-documented in AI bias research, even though this argument does not cite it. The weakness lies in asserting a specific theological mechanism ('total depravity,' 'unregenerate minds') as if it were established by, rather than merely compatible with, that empirical pattern, and in not explaining why this makes AI categorically different from other human artifacts.
• P4 (Scriptural sufficiency for evaluation) (Moderate) — Coherent as a corollary of A1, but the specific hermeneutical claim that Scripture alone (rather than Scripture plus natural revelation or reasoned reflection, as some Christian traditions hold) suffices to evaluate a technology as novel as AI is asserted rather than defended, and it offers no bridge from abstract principles (image of God, wisdom/folly) to concrete technical evaluation criteria.

Potential Fallacies

• False dichotomy (P1) — P1 asserts that moral evaluation either rests on a transcendent standard or collapses into 'whatever moral framework presently dominates the culture.' This excludes secular moral realism, natural law theory, Kantian deontology, contractarianism, and virtue ethics — traditions that claim non-relativistic normative grounding without appeal to revelation. Because this premise is the hinge connecting general meta-ethics to the specific claim that Scripture must be the standard, its unproven exhaustiveness is the argument's most load-bearing vulnerability. Functioning…
• Equivocation between theological and technical senses of 'bias' (P3) — P3 moves from the doctrine that human thought is corrupted by sin to the claim that AI systems will carry 'the biases... of unregenerate minds' at scale, treating this as though it were the same phenomenon documented in machine-learning bias research (training-data skew, labeling artifacts, optimization objectives). The empirical pattern is real and well-established, but the theological gloss adds interpretive weight that the technical evidence alone does not establish, without clarifying whether 'depravity' functions as…
• Category leap / potential special pleading (P3 / conclusion framing relative to premises) — The argument moves from a defensible claim (AI raises theological and ethical questions) to the stronger claim that AI specifically is a 'gospel' (soteriological) matter, without a principled criterion distinguishing AI from other human inventions that likewise 'bear the fingerprints of sinners' (printing press, spreadsheets, agriculture). Absent such a criterion, the 'gospel question' category risks becoming so expansive that it loses discriminating force, or risks appearing to single out AI for…

Counterarguments

• P1 (High impact) — Functioning, historically durable secular ethical frameworks — bioethics committees, engineering codes of conduct, international human rights law — demonstrate non-relativistic normative grounding without appeal to revealed theology, directly falsifying the claim that absent a transcendent standard, ethics necessarily collapses into whatever the culture currently favors.
• P3 (Medium impact) — Documented causes of algorithmic bias — unrepresentative training data, labeling practices, optimization objectives, feedback loops — provide a complete and empirically testable causal account of biased AI outputs without requiring appeal to a doctrine of total depravity, making the theological explanation additive rhetoric rather than necessary explanation.
• Conclusion / P3 (Medium impact) — If any artifact 'bearing the fingerprints of sinners' warrants gospel-level theological scrutiny, then virtually all human technology (language, mathematics, the printing press, spreadsheets) would qualify equally, diluting the category to the point of losing discriminating force; the argument does not supply a principled threshold for why AI crosses into 'gospel question' territory while other technologies do not.
• P4 (Medium impact) — Christian traditions themselves disagree about whether Scripture alone is sufficient for evaluating novel technologies or whether natural law and general revelation must supplement it; this internal disagreement undercuts the premise's claim to settled sufficiency.
• Conclusion (Low impact) — The practical recommendation — discernment rather than panic or uncritical embrace — is functionally indistinguishable from what careful secular ethicists already advise, suggesting the theological superstructure, while internally meaningful, is not action-guiding in ways that go beyond generic prudential ethics.

Suggested Improvements

• P1's meta-ethical dichotomy — Acknowledge and directly engage the strongest non-theistic candidates for objective moral grounding (natural law, moral realism, Kantian universalism) rather than treating the transcendent/relativist split as exhaustive. This is the argument's most load-bearing and most contested premise; addressing rival views would substantially strengthen its persuasive reach beyond an audience that already grants A1.
• P3's bias framing — Separate the empirically supported technical claim (AI encodes the biases of its data and designers) from the theological interpretation (this reflects total depravity), and cite concrete mechanisms (training data, RLHF, objective functions) as the bridge between the two. This would ground the theological claim in verifiable technical reality rather than leaving it as an assertion riding on the coattails of a separate empirical phenomenon.
• Distinguishing AI from other technologies — Articulate a specific criterion (e.g., scale, autonomy, mimicry of cognition, feedback with culture) that explains why AI crosses a theological threshold that simpler tools do not. Without this, the argument is vulnerable to a reductio that trivializes the 'gospel question' category by extending it to all human artifacts.
• Actionability — Supply worked examples or criteria translating 'image of God,' 'wisdom and folly,' and 'sin' into concrete evaluative questions for specific AI use cases. The current level of abstraction makes 'discernment' difficult to operationalize consistently across individuals or institutions, risking inconsistent or ad hoc application despite the claimed objectivity of the standard.
• Engagement with secular AI ethics — Reference or respond to existing secular frameworks (fairness/accountability/transparency literature, alignment research) rather than proceeding as if theological categories are the only lens addressing AI's encoded biases. This would demonstrate awareness of the broader AI ethics conversation and position the theological framework as a distinct contribution rather than an isolated parallel discourse.

Scenario Tests

• A secular bioethics board or engineering code of conduct successfully governs a technology's ethical use without appeal to any transcendent standard. (Challenges) — Directly undermines P1's claim that absent transcendence, ethics collapses into cultural dominance, since such institutions demonstrate durable non-relativistic normativity without revelation.
• The argument's logic (bears fingerprints of sinners, therefore a gospel question) is applied to a simpler technology such as the printing press or a calculator. (Challenges) — Exposes the special-pleading concern: without a stated threshold distinguishing AI, the same reasoning would classify nearly all technology as a gospel matter, diluting the conclusion's force.
• AI bias in a deployed system is fully diagnosed and corrected through technical means (data rebalancing, fairness audits) without any theological input. (Challenges) — Suggests P3's causal story is not necessary to address the phenomenon it describes, weakening the claim that only judgment by 'God's revealed will' can address AI's inherited biases.
• The argument is presented to an audience that already accepts Reformed evangelical theological commitments (biblical authority, total depravity). (Supports) — Within this audience, the argument functions coherently as an internal application of shared doctrine to a new domain, offering a a stable 'third way' between panic and uncritical adoption.
• The argument is presented in a pluralistic public policy context where AI governance decisions must be justified to stakeholders who do not share A1. (Challenges) — The argument's persuasive traction collapses outside its target audience, since its central claims about the necessity of Scripture as evaluative standard depend entirely on assumptions not shared in secular or interfaith settings.

Coherence & Relevance

The argument is internally coherent and free of classical formal fallacies once its stated assumptions are granted: each premise contributes to a convergent case that AI cannot be evaluated by usefulness alone and must be judged theologically. Its principal vulnerabilities are substantive rather than structural — an unproven exhaustive dichotomy in P1, an under-argued slide between theological and technical senses of bias in P3, and an unexplained threshold for why AI specifically (rather than technology generally) crosses into 'gospel' territory. The argument's force and persuasive reach are therefore highly audience-dependent: robust for those who already accept its Reformed evangelical starting points, and substantially weaker for audiences who do not share those foundational commitments.

• P1 (meta-ethical collapse without transcendent standard) (Strong) — Directly motivates the need for a transcendent standard, but its exhaustiveness is unproven; if secular non-relativist frameworks are viable, the inferential link from 'AI raises normative questions' to 'therefore Scripture is required' weakens substantially.
• P2 (universal sinfulness and 'many devices') (Strong) — Well-connected to the conclusion's claim that AI cannot be neutral, though it largely restates assumption A1 in scriptural terms rather than adding independent evidential weight.
• P3 (AI mimics reasoning, inherits biases at scale) (Moderate) — Provides the crucial link explaining why AI specifically (not just technology generally) warrants theological scrutiny, but conflates an empirically supported technical claim with an asserted theological mechanism, and does not establish a clear threshold distinguishing AI from other artifacts.
• P4 (Scriptural sufficiency) (Moderate) — Supports the practical conclusion (discernment, not fear) but leaves a gap between abstract scriptural principles and concrete technical evaluation, and assumes a sufficiency doctrine not universally held even within Christian thought.

View this argument on LogicFirst.ai