Original Research

Imago Dei in the age of algorithms: A theologically grounded critique of artificial intelligence in Christian education

Daniel Sudibyo Tjandra

Received: 23 Mar. 2026; Accepted: 03 Aug. 2026; Published: 14 Sept. 2026

Copyright: © 2026. The Author Licensee: AOSIS.
This work is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license (https://creativecommons.org/licenses/by/4.0/).

Abstract

The embedding of artificial intelligence (AI) in education confronted Christian institutions with a challenge technical adjustment could not resolve, and existing theological responses had not mobilised the imago Dei as a systematic criterion for evaluating the anthropology AI presupposes. Christian schools, colleges and seminaries adopting adaptive learning platforms, automated assessment engines and language model tutors formed the setting, addressed within the reformed-ecumenical tradition. A qualitative, systematic-theological method combined critical-hermeneutical analysis of primary sources with normative argumentation. The argument was framed as normative rather than empirical. The study established a three-layered account of the imago Dei (vocational, relational and eschatological) and a two-axis typology distinguishing assistive from substitutive deployment and instrumental from formative domains. Theological objection attached specifically to substitutive-formative deployment, while assistive-instrumental deployment proved permissible. A theologia crucis and pneumatological pedagogy yielded a procurement protocol, protected practices and curricular proposals. Christian education requires an ontological criterion, not merely an ethical checklist, to discern which AI deployments serve and which corrode the formation of persons.

Intradisciplinary and/or interdisciplinary implications: This article brings systematic theology and theological anthropology into critical dialogue with AI ethics, philosophy of technology and Christian religious education, arguing that adequate responses to AI in education require theological depth joined to empirically grounded social critique.

Keywords: imago Dei; artificial intelligence; theologia crucis; pneumatological pedagogy; Christian education; theological anthropology; algorithmic bias; kenosis.

Introduction

The emergence of artificial intelligence (AI) as educational infrastructure, adaptive learning systems, automated assessment engines and large language model tutors, signifies more than a technological shift. It marks a reorientation of the educational paradigm and of the understanding of human identity that education presupposes (Selwyn 2019:3). Within a single generation, institutions once centred on the encounter between human teachers and students have been reorganised around computational systems that process learners as behavioural data, optimise instructional pathways through algorithmic inference and assess human development through quantifiable outputs (Crawford 2021:1). The central issue extends beyond pedagogical effectiveness to the anthropological imagination that shapes the entire educational enterprise. For Christian educational institutions, this challenge carries implications that secular frameworks are structurally unable to address in full. The Christian tradition regards education as fundamentally anthropological: teaching, learning, discipleship and formation are rooted in, and accountable to, a particular understanding of human nature and purpose (Middleton 2005:25). The imago Dei, the affirmation that human beings are created in the image of God, is the central concept guiding this vision (Gn 1:26–27). This conviction, which has shaped Christian reflection on human dignity and vocation from the patristic era to the present, functions as the normative theological criterion against which any educational proposal must be measured. The critical question is not whether AI can technically support certain functions, but whether the anthropological logic underlying particular AI deployments can be reconciled with the Christian understanding of the human person. That question has become urgent because a significant tension exists between the computational ontology on which AI systems operate and the imago Dei tradition. Artificial intelligence systems reduce complex phenomena, including human beings, to formal representations that can be quantified and optimised (Coeckelbergh 2020:24). When an AI system processes a student’s learning behaviour, it necessarily represents the student as a data-generating agent rather than as a person whose dignity and eschatological vocation exceed any formal representation. Whether this representational reduction becomes an anthropological reduction, however, depends on how the system is deployed, a distinction this article develops at length. Recent scholarship has begun to address this challenge within faith-based educational traditions, and it has done so from more than one direction. Corpuz (2025:2) illustrates how generative AI can enhance reflection in Catholic theological classrooms when guided by theological virtue ethics. Papakostas (2025:12) concludes from institutional case studies that AI can enrich religious education when deployed thoughtfully, provided it does not displace the relational and formational dimensions at the heart of the discipline. Hwang (2026) examines how AI transforms the conditions for grace, community and religious formation in digital educational environments through a Wesleyan lens. Working in the opposite direction, Molhoek (2022:4) and Okeke (2025:9) explore whether AI might itself be understood as bearing something of the divine image mediated through human creativity, while Auh and Kim (2025:6) treat the imago Dei as a positive source of governance norms for AI design. Hermanto (2025:5) contrasts human personhood, grounded in Trinitarian relationality, with what he terms the ‘simulated personhood’ of AI systems. These contributions are substantial, yet none has developed the imago Dei tradition systematically, across its vocational, relational and eschatological dimensions, as a criterion for discriminating between theologically acceptable and theologically corrosive deployments of AI in Christian education. Nor has a theologically robust account that begins with the imago Dei, incorporates the theologia crucis and culminates in a constructive pneumatological pedagogy yet been offered. This study addresses that gap. This study aims to develop a theologically grounded critique of AI in Christian education by constructing the imago Dei as a comprehensive criterion of ontological resistance to anthropological reductionism, by anatomising the mechanisms of algorithmic distortion with theological precision, by distinguishing the deployments of AI to which theological objection properly attaches from those to which it does not, and by proposing the theologia crucis together with a pneumatological pedagogy rooted in the Spirit’s formative work in community as a constructive alternative for Christian educational practice. This study is a qualitative, systematic-theological inquiry that combines critical-hermeneutical analysis of primary theological sources with normative constructive argumentation. It follows the pattern of doctrinal-critical research in which a classical theological locus is retrieved, developed and then brought to bear as an evaluative criterion on a contemporary practice. The argument is normative rather than empirical, and the distinction matters for how its claims should be assessed. The study does not assert, and its method could not establish, that every AI system necessarily produces anthropological reduction in every educational setting. Its claims are of a different order: that certain deployments of AI in education instantiate an anthropology incompatible with the imago Dei tradition; that this incompatibility is a matter of the operative logic of those deployments rather than of implementation quality; and that Christian educational institutions therefore require theological criteria, not merely ethical safeguards, for discerning between them. Documented cases are adduced as illustrations of what the normative analysis identifies, not as an empirical demonstration of necessity. Because ‘ontological resistance’ can be read in several ways, its sense here requires precision. Three distinct claims must be distinguished. The first is a metaphysical claim: that AI is incapable in principle of bearing the divine image. This study does not argue that. Whether an artificial general intelligence could be an image-bearer is a live question in the literature (Molhoek 2022:8), and it is not this article’s question. The second is an anthropological-descriptive claim: that the account of the human person operative within certain AI systems, the person as a data-generating agent whose development is predictable, optimisable and assessable through quantified outputs, is incompatible with the account of the person given in the imago Dei tradition. This study does argue that, and it is the core of the analysis. The third is a practical-normative claim: that Christian educational institutions have a theological obligation to refuse deployments of AI whose operative logic instantiates the incompatible anthropology, and to structure permissible deployments so that the theological account of the person governs the technology rather than the reverse. This study argues that too, and derives concrete practices from it. ‘Ontological resistance’, then, names the second and third claims taken together: a resistance grounded in a rival account of what a human being is, issuing in specific institutional practice. It is neither a metaphysical verdict on machine capacity nor merely an ethical objection to particular harms.

Person before the algorithm: Imago Dei as ontological resistance

Framing the imago Dei as ontological resistance treats this theological concept as deliberately polemical: A contest between rival accounts of what a human being is. The imago Dei resists reductive anthropology not from defensive anxiety but through the strength of an ontological claim that precedes, exceeds and judges every human cultural configuration, including the algorithmic. This section demonstrates why the doctrine constitutes a formidable resistance to reductive anthropological logic, and argues that allowing it to remain at the level of pious affirmation surrenders the most powerful theological resource available to Christian educators. The history of interpretation presents multiple trajectories, not all equally adequate to the present challenge. The substantialist reading, dominant from Augustine through Aquinas, locates the image in a specific cognitive capacity, most commonly the intellect (Hoekema 1986:42). While this reading affirms rational dignity, it is dangerously partial. By centring the image in cognitive function, it inadvertently provides a foothold for the very anthropology it should resist: if the image is equated with the highest form of rational function, and AI systems increasingly demonstrate sophisticated rational capacities, then maintaining a principled distinction between human dignity and machine capability becomes difficult within a purely substantialist framework. A more robust account requires the tradition’s functional and relational trajectories. Middleton (2005:89) develops a rigorous functional interpretation through attention to ancient Near Eastern royal ideology, identifying the imago Dei not as a static cognitive quality but as a royal vocation: human beings are commissioned as God’s image-bearing representatives to exercise responsible stewardship over creation in a manner reflecting God’s own redemptive purposes. The implications are decisive. Artificial intelligence systems do not exercise vocation; they execute functions. The difference is ontological rather than merely semantic. A vocation is a calling received from a personal God, lived out in responsible, temporally extended agency oriented towards a divinely given telos. A function is a computational operation defined by inputs, outputs and optimisation criteria. Gunton (1998:9) deepens this insight: human beings, as images of the God who creates freely and purposefully, are characterised by a freedom and purposiveness that no computational model captures. The relational interpretation, developed extensively by Barth, introduces the dimension of interpersonal constitution. For Barth (1960:182), the image is not a property of the human being in isolation but a relational category: the human being is constituted as image-bearer through encounter [Gegenüber] with the divine Thou and the human other, reflecting the eternal self-differentiation and self-giving within the Trinity. Grenz (2001:141) grounds this relational reading in Trinitarian social ontology, arguing that human beings reflect the image of the Triune God not as isolated individuals but as persons-in-community whose relational existence mirrors the perichoretic communion of Father, Son and Spirit. Zizioulas (1985:27) further asserts that personhood, unlike mere individuality, is constituted through loving self-donation in community, and Volf (1998:178) argues that genuine personhood requires a mutual indwelling that algorithmic mediation does not reproduce. The eschatological dimension, often neglected in discussions of AI, supplies a final and perhaps most penetrating layer. Pannenberg (1985:208) contends that human beings are constituted not only by their origin in creation, but also by their destiny in Christ: the full realisation of the image-bearer is proleptic, situated in the eschatological future in which the Spirit conforms the person to the image of the Son (Col 3:10; Rm 8:29). This eschatological constitution implies that the truth about a human being cannot be discerned from present behavioural data. A student profiled by an AI system through interaction analytics is not a completed reality but a person-in-becoming, whose truest identity is defined by divine promise rather than algorithmic prediction. Artificial intelligence processes what a person presently is; it has no access to what a person shall be. The stewardship mandate embedded in the doctrine supplies a crucial constructive complement to these critical dimensions. Herzfeld (2002:75) contends that the image-bearing vocation of responsible stewardship over creation, including technological products, requires Christian engagement with AI to be characterised by critical oversight and accountable governance rather than passive adoption or reflexive rejection. This is an important qualification. The imago Dei affirms that human beings are the subjects of the technological order, not its objects, which entails responsibility for shaping technology, not withdrawal from it. Christian educational institutions therefore bear a theological responsibility to evaluate every AI system against the criterion of the imago Dei, and to refuse deployment of any system whose operative logic contradicts the tradition’s multi-dimensional account of personhood. Taken together, the vocational, relational and eschatological dimensions establish a three-layered resistance. At the vocational layer, the image-bearing person is distinguished by the irreducibility of vocation, a calling that cannot be optimised or automated. At the relational layer, the person is distinguished by the constitutive significance of genuine I–Thou encounter, an address that no simulation replicates. At the eschatological layer, the person is distinguished by openness to a divinely promised future identity that exceeds any behavioural data profile. Collectively these layers constitute a theological account of personhood that resists reductionism and provides a generative framework for practice, clarifying what Christian education must protect and cultivate in a world increasingly shaped by algorithmic systems.

When data replaces the soul: A theological anatomy of artificial intelligence’s anthropological reductionism

A theological anatomy, understood as the disciplined dissection of a pathology, is more useful here than polemical denunciation. The anthropological danger does not lie in overtly malign systems, which are readily identified and rejected. It is found in a subtle and systemic reductionism, often presented under the guise of improvement: more personalised learning, more efficient assessment, more equitable access. Analysis must penetrate these appearances to reveal the mechanisms by which particular deployments distort the human person in ways that are both documentable and theologically significant. Floridi’s (2014:46) concept of the infosphere offers a useful framework: human beings risk being reconstituted as informational organisms whose identity is defined by their data profile rather than by their ontological status as image-bearers. The most immediate mechanism is algorithmic objectification: the systematic conversion of the lived, irreducible experience of the learner into quantifiable data points suitable for computational processing. Turkle’s (2015:3) analysis of the losses incurred when digital mediation replaces face-to-face encounter highlights the phenomenological dimension of this distortion. Substitutive tutoring systems are structurally incapable of supplying the most formationally significant qualities of human relationship – sustained attention, empathic responsiveness and willingness to be changed by the other. In such deployments the learner is addressed not as a person whose complexity requires attentive engagement but as a user profile whose predicted behaviours can be optimised through targeted content delivery. From the perspective of the relational imago Dei, this is not merely a pedagogical inconvenience but an anthropological crisis. O’Neil’s (2016:52) examination of algorithmic systems in high-stakes domains demonstrates that reductive objectification is not an incidental flaw of early systems but a structural feature of the optimisation logic underpinning them. Predictive models used for assessing student risk, ranking teacher performance and screening applicants operate according to a common logic: they reduce the complex, historically embedded and relationally constituted reality of persons to variables processed by optimisation functions, and they routinely disadvantage students from marginalised communities, encoding historical inequalities into predictions of future performance under the guise of mathematical objectivity. Noble (2018:124) extends this critique by demonstrating that such systems reflect racial and class hierarchies systematically rather than accidentally, while Eubanks (2018:11) introduces the concept of automated inequality, in which algorithmic systems surveil and control marginalised populations, encoding poverty as pathology. Crawford (2021:211) situates these harms within the political economy of AI’s extraction from communities, and Benjamin (2019:15) offers a particularly incisive diagnosis in the ‘New Jim Code’: the reproduction of racial hierarchy through systems that appear race-neutral. The 2020 English school qualifications episode illustrates precisely what these analyses describe. When examinations were cancelled during the COVID-19 pandemic, the Office of Qualifications and Examinations Regulation replaced teacher-assessed grades with grades calculated by an algorithm that combined each school’s historical grade distribution with within-school student rankings. Approximately two-fifths of teacher assessments were downgraded. Because the model weighted institutional history heavily, high-achieving students in historically lower-attaining state schools were systematically downgraded while students in small classes at well-resourced independent schools were comparatively advantaged; students from disadvantaged backgrounds were hit hardest, and the regulator’s own equalities analysis had failed to detect the effect (Smith 2020:1077). The grades were withdrawn after public protest. The episode is instructive theologically as well as politically: an individual student’s eschatological openness – the possibility of becoming other than her institution’s history predicts – was precisely what the model was designed to suppress in the name of consistency. What Pannenberg identifies as constitutive of the person was treated as statistical noise. Comparable dynamics are documented in early-alert retention analytics, automated essay scoring and remote proctoring systems, each of which converts a dimension of the learner into a tractable variable. The theological assessment of these distortions must be diagnostic and prophetic. Moltmann’s (1985:216) account of human dignity, rooted in the relational constitution of the person as image-bearer, grounds the diagnosis: when algorithmic systems reduce individuals to data profiles, they perpetrate ontological violence against the irreducible personhood of those created in God’s image. Hauerwas (1981:89) further contends that human beings are narrative creatures whose identities are shaped by the stories they inhabit and the communities in which they are formed. Educational systems governed by algorithmic processing sever learning from the narrative and communal contexts that give it a distinctively human character, producing competency profiles rather than educated persons. The prophetic dimension is grounded in Scripture’s persistent defence of the structurally vulnerable against systems of power that deny their dignity. The oracles of Isaiah, Amos and Micah constitute a theological imperative to expose and resist such systems, and that imperative applies with full force to algorithmic arrangements that disadvantage the already disadvantaged (Is 1:17; Am 5:24; Mi 6:8). Vallor’s (2016:21) philosophical account of the technomoral virtues, such as humility, care, honesty, justice and practical wisdom, identifies the dispositional qualities Christian educators must cultivate to exercise this prophetic vocation, and Corpuz (2025:7) confirms that justice, honesty, responsibility and care constitute the foundational virtues through which AI can be held accountable in theological educational settings.

Counter-positions: The imago Dei as a resource for design

A significant body of scholarship reads the relationship between the imago Dei and AI more constructively, and the argument advanced here must be tested against it. Three positions deserve particular attention. Molhoek (2022:8), working from Hefner’s ‘created co-creator’ model, asks whether artificial general intelligence produced by genuinely cooperative human creative action might be made in the image of humanity and, through that mediation, stand in some relation to the image of God. Okeke (2025:11) develops a parallel proposal, characterising AI as imago hominis, an expression of human creativity and therefore a legitimate development within creation rather than an intrusion upon it, and argues that theological scepticism should be moderated accordingly. Auh and Kim (2025:8) turn the doctrine in a different direction, treating the imago Dei not as a criterion of critique but as a positive source of governance norms, from which they derive requirements for bias auditing, algorithmic transparency and mandatory human oversight. Xu (2024) similarly mobilises the doctrine constructively for a theological account of digital technology in liturgical and ethical practice, and Platovnjak and Svetelj (2024:844) argue that AI’s principal theological significance lies in its capacity to provoke a rediscovery of what is distinctive about human nature. Two responses are warranted. The first is concessive. The governance proposals of Auh and Kim are not in tension with the present argument but complementary to it, and they are here endorsed: the stewardship dimension of the imago Dei, developed above with Herzfeld, generates exactly such obligations. Where the imago Dei informs design and governance, it functions constructively; the present critique concerns what happens when it is absent from deployment decisions. Przygoda, Rynio and Kalisz (2025:14) draw the correct pastoral conclusion when they argue that the churches’ task is not to prohibit the faithful from using these technologies but to teach their proper use, a formative rather than prohibitive posture that the typology developed above is designed to make operable. The second response is delimiting. The created co-creator and imago hominis proposals address a different question from the one pursued here. They concern the ontological status of the artefact; this article concerns the anthropological status of the learner. One may grant that AI is a legitimate product of image-bearing human creativity – indeed the stewardship mandate suggests as much – while still maintaining that a particular deployment of that artefact misrepresents the students it processes. An artefact may be theologically licit in its origin and corrosive in its application. The critique developed here is therefore compatible with Molhoek’s and Okeke’s conclusions rather than opposed to them, once the two questions are distinguished. Hermanto (2025:9) marks the relevant boundary precisely: whatever is granted about AI as artefact, the ‘simulated personhood’ it presents to a learner is not the relational personhood constituted by participation in the Spirit, and a formation programme that treats the simulation as sufficient has substituted the sign for the reality. This analysis reveals a pattern of distortion operating at multiple levels: ontological, in the reduction of persons to data; relational, in the replacement of encounter with interface; structural, in the perpetuation of injustice through bias; and narrative, in the fragmentation of the self from its formative community. Coeckelbergh’s (2020:24) observation that AI ethics has largely operated within thin philosophical frameworks is substantiated by this analysis. A robust, theologically grounded anthropology drawing on the full resources of the imago Dei tradition is required to comprehend what is at stake when data threatens to supplant the soul.

Theologia crucis as a pedagogical counter-proposal: Kenosis, vulnerability and the optimisation refusal

A theologically robust response emerges not only from the imago Dei but, more profoundly, also from the theologia crucis. The educational logic of algorithmic optimisation exemplifies what Luther identified as the theologia gloriae: a mode of knowing and acting that prioritises power, clarity, efficiency and the elimination of weakness and ambiguity. Christian education, by contrast, is rooted in a theology that locates revelation in weakness, formation in vulnerability and the deepest understanding in embracing what the logic of glory rejects. This opposition supplies a critique that is not only ethical or anthropological but epistemological, addressing the structures of knowledge and formation that shape Christian education. Luther’s Heidelberg Disputation provides the foundational articulation. In theses 19–21, he draws a sharp distinction between the theologian of glory, who seeks God through visible majesty, power and wisdom, and the theologian of the cross, who perceives God in suffering, weakness and the apparent folly of the cross (1 Cor 1:18–25; Luther 1989:48). The theologian of glory misapplies human categories of achievement and thereby misinterprets both God and the human person. Educationally, optimisation that eliminates uncertainty and quantifies outcomes is the pedagogical manifestation of the theology of glory: it seeks clear, measurable results by removing the vulnerability and open-endedness essential to authentic human formation. Moltmann (1974:65) extends the Lutheran theologia crucis into a systematic framework with significant educational implications, arguing that the God revealed in the cross is not defined by power and impassibility but suffers with and within created reality, incorporating the vulnerability and contradictions of human existence into the divine life itself.

The deepest theological truth is therefore found not in optimisation but in kenosis and not in the elimination of weakness but in its redemptive embrace as the locus of God’s transformative presence. An educational process that avoids engaging students in their vulnerability, treating error solely as an efficiency problem to be corrected by algorithmic means rather than as a meaningful moment of growth, is not merely pedagogically inadequate but theologically inconsistent with the God revealed in the crucified Christ. The Pauline theology of kenosis establishes the Christological framework underpinning this educational paradigm. The self-emptying of the pre-existent Son, the voluntary surrender of divine prerogatives to assume the form of a servant and embrace human vulnerability, is not only a salvific event but a pattern for authentically Christ-shaped human existence (Phlp 2:5–8). Pedagogically, the Christian educator is called to a kenotic stance: relinquishing control, engaging authentically with the learner’s confusion and struggle, being transformed by the encounter and prioritising genuine development over the production of measurable outcomes. Hall (1976:115) affirms this pedagogical kenosis in his reading of the theologia crucis as a response to the triumphalism of technological culture, asserting that authentic Christian witness requires willingness to occupy the position of weakness that a culture of algorithmic glory consistently rejects. Biesta (2014:1) provides a secular parallel that is instructive precisely because of its independence from theological frameworks. Authentic education, he argues, is inherently risky, requiring educators to engage students as genuine others whose responses cannot be predicted, controlled or optimised. This ‘beautiful risk’, relinquishing control to foster genuine subjectivity, corresponds theologically to kenosis. Substitutive systems are designed to avoid exactly this risk: they minimise variance, reduce uncertainty and optimise for predicted outcomes, eliminating the conditions Biesta identifies as essential to education and that the theologia crucis regards as foundational to formation. Postman (1992:13) describes the broader cultural context in his account of technopoly, in which technology becomes the ultimate authority and supplants moral and humanistic criteria with technical efficiency. Freire’s radical pedagogy resonates powerfully here. His ‘banking’ concept of education, in which the teacher deposits knowledge into the passive student, is the educational equivalent of the theologia gloriae: power, knowledge and the elimination of the learner’s agency organised around the optimising control of the educator or the system (Freire 1970:71). Genuine pedagogy requires the educator to relinquish the position of omniscient authority and engage in co-investigation with the learner, a posture structurally identical to kenosis. Hauerwas (1981:125) further asserts that suffering and vulnerability form the deepest dimensions of moral character, and that the most significant formation occurs in the experience of limitation and the need for community. A system that algorithmically manages learner experience to minimise these uncomfortable moments impedes the formation of the character Christian education seeks to cultivate. The cruciform pedagogy that emerges does not reject educational structure. It constitutes a disciplined counter-cultural practice grounded in the conviction that the deepest truths about persons and God are disclosed in weakness rather than power, and in genuine encounter rather than optimisation. MacIntyre (1981:191) supplies a philosophical complement in his argument that practical wisdom cannot be reduced to algorithmic rules but requires character formed through sustained participation in communities with a shared telos. The cruciform educator is the practitioner of phronesis in the domain of Christian formation, whose wise, attentive and kenotic engagement with students is cultivated not by algorithmic training but by the community of the cross, the practices of prayer and worship, and the slow, patient formation of virtue in the Spirit.

Pneumatological pedagogy: Formation, community and limits of the algorithmic teacher

The preceding analysis culminates in a pneumatological pedagogy: a model of Christian educational practice that recognises the Holy Spirit as the primary and irreplaceable Teacher within the Christian community, in which human educators participate in the Spirit’s formative work through relational, communal, embodied and ecclesial practices. This proposal is anchored in the Johannine promise of the Paraclete, whom the Father will send to teach the community everything (Jn 14:26); a lasting theological claim that the deepest Christian teaching resides not in external systems but in the internal, communal and relational work of the Spirit, who searches the depths of God and reveals them to the community of faith (1 Cor 2:10–13). Fee (1994:5) provides a precise pneumatological foundation. For Paul, the Spirit is the primary agent of formation and the renewal of the mind within the community of believers (Rm 12:2). Formation in the image of Christ occurs through communal practices – prayer, Scripture reading, worship, the Lord’s Supper, mutual exhortation and accountability – that require embodied and relational presence. The Spirit’s formative work is consequently community-dependent, and AI systems cannot participate in this dynamic because they are not members of the faith community: they do not pray, worship, bear burdens or receive the Spirit’s formation. This is the theological ground of the substitutive-formative prohibition proposed above. Yong (2005:29) introduces a crucial further perspective: the Spirit poured out on all flesh (Ac 2:17) forms the faith community as a radically inclusive and irreducibly diverse body in which each member contributes essential gifts. This vision of communal diversity stands in sharp contrast to the homogenising tendencies documented by Noble, Eubanks and Benjamin, whereby systems encode dominant cultural patterns into their outputs.

A pneumatological pedagogy attends to the particular formation of particular persons within particular communities, resisting the reduction of human diversity to data patterns. Pinnock (1996:9) clarifies why simulated relationship is not merely inadequate but counterfeit: the Spirit is the relational power of divine love that renders God’s presence concretely personal and responsive, and a simulation offers the appearance of personal presence without its reality. Groome (1980:184) offers the most comprehensive account of pedagogy structured around pneumatological formation in practice. His five-movement structure of shared Christian praxis, naming present praxis, critical reflection, the Christian Story and Vision, appropriation and decision for transformed praxis, engages the whole person in communal, dialogical, Spirit-led encounter with the living tradition. The human educator is indispensable within this framework: shared praxis requires a human presence capable of listening, responding, challenging, discerning, praying and accompanying. Artificial intelligence may support the process peripherally by supplying informational resources and widening access to diverse perspectives, but it cannot occupy the central role without altering the character of the praxis encounter. Westerhoff (1976:78) asserts that faith is not a cognitive achievement resulting from instruction but a quality of life shaped through sustained participation in communal practices, such as worship, service, witness, lament and the apprenticeship of belonging to a community with a shared narrative. Pazmiño (2008:175) offers a complementary six-dimensional account of Christian formation, encompassing cognitive, affective, behavioural, spiritual, communal and prophetic dimensions. Artificial intelligence may address the cognitive and behavioural dimensions to some extent; the spiritual, communal and prophetic dimensions require the embodied, Spirit-empowered presence of persons in community. Bosch (1991:390) reinforces the point: the faith community as the hermeneutic of the gospel is itself the irreplaceable outcome of pneumatological formation. The Wesleyan tradition supplies distinctive resources for articulating these limits. The Wesleyan means of grace – prayer, Scripture, fasting, holy conferencing and the Lord’s Supper – are inherently communal and embodied practices through which the Spirit forms persons in holiness, and digitally mediated educational environments must therefore be designed to preserve and extend these practices rather than to attenuate them (Hwang 2026). The work of Hooks (1994:13) on education as the practice of freedom, in which students are liberated from oppressive systems to engage their critical consciousness, offers a secular analogue to the Spirit’s liberating work, while Corpuz (2025:9) demonstrates that AI can serve as a catalyst for theological reflection and virtue formation when integrated into a framework centred on human encounter, communal discernment and critical engagement.

From principle to practice: A procurement and formation protocol

Because such proposals commonly remain at the level of principle, the following concrete practices are offered for institutional adoption.

A theological triage protocol for procurement

Before adoption, each proposed system should be located on the two-axis typology set out above. Assistive-instrumental tools proceed under normal governance. Assistive-formative tools require a written theological justification identifying which formative function is mediated, what human oversight is retained and how the tool’s implicit anthropology has been assessed. Substitutive-formative tools are declined. Responsibility for this determination should rest with a standing body including theological faculty, not with procurement or information-technology staff alone.

Protected practices

Institutions should name explicitly, in policy, those practices that are not subject to algorithmic mediation: corporate worship, confession and absolution, spiritual direction, lament, pastoral crisis response, communal discernment and the mentoring relationship between teacher and student. Naming them in advance prevents their erosion by incremental efficiency decisions.

Assessment redesign

Because the theologia crucis identifies risk and open-endedness as constitutive of formation, assessment should be redesigned to protect them: viva voce examination, portfolio assessment that documents process rather than product alone, supervised theological reflection and communal presentation before peers. These forms are resistant to substitution precisely because they require presence.

Reinvestment of released time

Where assistive tools reduce administrative burden, institutions should specify how the released time is reallocated to relational contact – tutorial hours, spiritual direction, pastoral availability – and should measure this reallocation. Without such specification, efficiency gains are absorbed by workload intensification rather than converted into formation.

Curricular artificial intelligence literacy

Christian education curricula should incorporate a module addressing the technical operation of machine learning systems, documented cases of algorithmic bias, the theological anthropology implicit in optimisation logic and practices of discernment for personal use. Literacy of this kind equips students for faithful agency rather than either uncritical adoption or uninformed refusal.

Formation of educators

Institutions should provide continuing formation for teaching staff in kenotic pedagogy: Supervision practices that build tolerance for productive confusion, mentoring in the accompaniment of struggling students and communal reflection on the temptation to reach for efficiency where presence is required.

Communal discernment of policy

Institutional AI policy should be developed through processes of communal discernment, faculty, students, chaplaincy and governance together, that explicitly invoke the theological criteria named here, rather than being adopted from sector templates.

Conclusion

This article has argued that the intersection of AI and Christian education requires a response that is ontological, prophetic, soteriological and pneumatological. Ontologically, the imago Dei, understood in its vocational, relational and eschatological dimensions, offers a comprehensive criterion for resisting anthropological reductionism. Prophetically, the mechanisms of algorithmic objectification, bias and structural injustice must be identified and addressed with the urgency of Scripture’s prophetic witness. Soteriologically, the theologia crucis presents a distinctively Christian alternative to optimisation logic, advocating kenosis and vulnerability as essential to authentic formation. Pneumatologically, the Spirit’s formative work within the faith community establishes the theological limit of algorithmic mediation. The argument has been deliberately delimited. Its objection attaches to substitutive deployment within formative domains rather than to educational AI as such, and it is compatible with, indeed dependent upon, the constructive proposals of those who mobilise the imago Dei for AI governance and design. What it insists upon is that the discrimination between permissible and corrosive deployment requires an ontological criterion and not merely an ethical checklist. Future research should test the procurement protocol proposed here in institutional settings, apply the theologia crucis to curricular design in Christian higher education, compare how different Christian traditions mobilise their theological resources in response to AI, and empirically assess pneumatologically informed pedagogical practices in AI-mediated environments.

Acknowledgements

The author declares that no financial or personal relationships inappropriately influenced the writing of this article.

Daniel Sudibyo Tjandra: Conceptualisation, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Visualisation, Writing – original draft, Writing – review & editing. The author confirms that this work is entirely their own, has reviewed the article, approved the final version for submission and publication and takes full responsibility for the integrity of its findings.

Ethical clearance to conduct this study was obtained from the Sekolah Tinggi Teologi IKAT Research Ethics Committee (No. 160/A/STTIKAT/II/2026).

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The author confirms that the data supporting this study and its findings are available within the article and its listed references.

The views and opinions expressed in this article are those of the author and are the product of professional research. They do not necessarily reflect the official policy or position of any affiliated institution, funder, agency or that of the publisher. The author is responsible for this article’s findings and content.

References