Epistemic Automation and the Deformation of the Human: Artificial Intelligence and the Reconfiguration of Theological Anthropology

Abstract

This paper argues that the most significant challenge artificial intelligence poses to theological anthropology is not ontological but epistemic. Rather than asking whether machines can think, feel, or bear the image of God, this paper redirects attention to the prior question of what happens to the human when core epistemic capacities, judgment, discernment, interpretive authority, and moral reasoning are progressively delegated to computational systems. Drawing on the concept of epistemic automation, understood as the systematic transfer of knowledge-producing functions from human agents to algorithmic processes, this paper develops a threefold analytical framework. First, it distinguishes epistemic authority from ontological status as the more productive locus for theological anthropological inquiry. Second, it introduces the distinction between fluency and understanding as an anthropological boundary condition that AI renders newly visible. Third, it analyses delegated cognition as a form of agency deformation with theological significance. The paper concludes that theological anthropology must move beyond reactive commentary on AI and instead generate a theory of the human under conditions of epistemic transformation. The argument engages constructively with philosophy of technology, social epistemology, and Christian theological traditions to offer a framework applicable across confessional boundaries.

1. Introduction

The encounter between artificial intelligence and theology has generated a rapidly growing literature, much of which is organised around a central question: what does it mean to be human if machines can replicate or surpass human cognitive performance? This question, while natural, has tended to channel theological reflection into a reactive posture. Theologians respond to developments in computer science, cognitive psychology, and philosophy of the mind by defending, revising, or abandoning particular accounts of human uniqueness. The result has been a body of work that is responsive but not yet architecturally ambitious, one that comments on AI rather than using AI as an occasion to reconstruct the conceptual foundations of theological anthropology itself.

This paper proposes a different approach. Rather than beginning with the ontological question, can machines think, feel, or bear the imago Dei, it begins with an epistemic question: what happens to the structure of human agency when the capacities that have traditionally grounded theological accounts of the person, judgment, discernment, interpretive authority, and moral reasoning are progressively delegated to algorithmic systems? The shift from ontological to epistemic framing is not a retreat from substantive theological questions. Rather, it is a reorientation that allows theological anthropology to engage the actually operative transformations of contemporary life, rather than speculative scenarios about machine consciousness that remain empirically underdetermined.

The concept that organises this reorientation is epistemic automation: the systematic transfer of knowledge-producing, knowledge-validating, and knowledge-applying functions from human agents to computational processes. Epistemic automation is not merely a technological phenomenon but an anthropological one, insofar as it alters the conditions under which knowledge is produced and validated. This shift has been analysed in philosophy of technology as a transition from tool-based cognition to system-level epistemic mediation (Alvarado 2023; Humphreys 2004; Coeckelbergh 2025). When a medical diagnosis is deferred to an algorithm, when a judicial risk assessment is generated by a predictive model, and when a student’s essay is evaluated by a language model, something happens not only to the task but to the person who would otherwise have performed it. The capacity that is delegated does not simply relocate; it atrophies, transforms, or is redefined in the agent who delegates it. This paper argues that such transformation constitutes a form of agency deformation that is theologically significant.

The argument proceeds in four stages. Section 2 surveys the existing landscape of theological engagement with AI and identifies the dominance of ontological framing as a structural limitation. Section 3 develops the concept of epistemic automation and argues that epistemic authority, rather than ontological status, is the more productive locus for theological anthropological inquiry. Section 4 introduces the distinction between fluency and understanding as an anthropological boundary condition that AI makes newly visible, and analyses delegated cognition as a deformation of agency with theological consequences. Section 5 draws constructive implications for the reconstruction of theological anthropology under conditions of epistemic transformation.

2. The Ontological Trap: AI and Theological Anthropology’s Reactive Posture

Theological engagement with artificial intelligence has largely been structured around a question of ontological comparison, namely whether machines possess, approximate, or fail to instantiate capacities traditionally associated with the human person. This comparative framing is widely observable across contemporary theological literature on AI (Dorobantu 2022; Graves 2024; Slattery and Green 2024). The dominant analytical structure is contrastive: human and machine capacities are placed side by side and evaluated relative to specific theological criteria, most prominently the doctrine of the imago Dei.

Within this framework, substantive accounts of the imago Dei, which locate the divine image in specific human capacities such as rationality, moral agency, or creativity, have come under pressure from advances in artificial intelligence. Systems capable of performing highly complex cognitive tasks in narrowly defined domains challenge the exclusivity of these capacities as uniquely human (Coeckelbergh 2020; Dorobantu 2024). In response, some theological approaches have shifted toward capacities that remain resistant to computational replication, including consciousness, subjective experience, and embodiment (Herzfeld 2002). Others have moved toward relational accounts, in which the imago Dei is defined in terms of the human capacity for communion with God and others, thereby relocating the theological boundary away from cognitive performance (McFadyen 1990). Functional accounts, by contrast, identify the imago Dei with vocation, such as stewardship or creative participation in the world, and have therefore opened the possibility that artificial systems might, in limited respects, participate in image-bearing functions (O’Donnell 2018). At the same time, this line of argument remains contested within theological anthropology. A number of theologians maintain that artificial systems, as artefacts of human design, are more appropriately understood as expressions of imago hominis rather than as participants in the imago Dei, thereby preserving a categorical distinction between human and artificial forms of agency (Herzfeld 2002; Waters 2006). This plurality of positions reflects a productive engagement with technological development, but it also reveals a structural limitation. The ontological framing constrains theological analysis to a reactive posture, in which each advance in artificial intelligence prompts a reassessment of human distinctiveness. As Dorobantu (2022) has argued, AI serves as a testing ground for theological anthropology, but the underlying logic of testing presupposes that the central question is whether machines meet or fail to meet human criteria. The result is a cycle of adjustment rather than the development of a stable theoretical architecture.

More importantly, the ontological question does not provide a mechanism for analysing how artificial intelligence reshapes the conditions under which human capacities are exercised. The question of whether a machine possesses a given property is distinct from the question of how the presence of such systems transforms the epistemic and social environments in which human agents operate. As Humphreys (2004) has noted in the context of computational science, technological systems can reconfigure the structure of inquiry itself, not merely extend existing capacities. The ontological question asks what a human is. The epistemic question asks how humans function under altered conditions of knowledge production.

It is this latter question that provides the point of departure for the present analysis.

3. Epistemic Automation: From Ontological Status to Epistemic Authority

3.1. The Concept of Epistemic Automation

The term epistemic automation refers to the systematic delegation of epistemic functions, including observation, interpretation, inference, evaluation, and judgment, to computational systems that execute these operations through algorithmic processes rather than through embodied human reasoning. This concept builds on work in philosophy of technology that characterises artificial intelligence as a fundamentally epistemic technology, that is, a class of artefacts designed to produce, manipulate, and validate knowledge claims (Alvarado 2023).

Epistemic automation must be distinguished from earlier forms of epistemic extension. Human beings have long relied on instruments that support cognition, including writing systems, measurement devices, and computational tools (Humphreys 2004). These technologies extend epistemic capacity while leaving interpretive authority primarily with the human agent. What distinguishes contemporary artificial intelligence is the relocation of interpretive and evaluative functions themselves. As Coeckelbergh (2025) argues, AI systems increasingly participate in belief formation and revision processes, thereby altering not only what is known but how knowledge is constituted.

The locus of interpretive authority has therefore shifted, such that human agents increasingly operate within distributed epistemic environments in which algorithmic outputs shape the conditions of judgment (Hauswald 2025; Fricker 2007). In social epistemology, this shift has been analysed in terms of artificial epistemic authorities. Hauswald (2025) examines the conditions under which artificial systems may be treated as sources of epistemic authority, that is, as entities with outputs that warrant a degree of deference. The emergence of such systems introduces a new layer into the epistemic landscape, one that operates independently of traditional criteria such as expertise, accountability, and social embeddedness (Fricker 2007).

3.2. Epistemic Authority Versus Ontological Status

A central analytical distinction follows from this development. Ontological status concerns the nature of a being, whether it qualifies as a person, agent, or bearer of the imago Dei. Epistemic authority concerns the capacity to produce knowledge claims that are treated as reliable and action-guiding within a given social context. These dimensions are analytically separable.

Contemporary artificial intelligence systems do not plausibly satisfy the criteria for personhood or image-bearing as understood within the Christian theological tradition (Herzfeld 2002; Waters 2006). However, they increasingly function as de facto epistemic authorities in practical contexts. Their outputs guide medical diagnoses, inform legal decision-making, structure communication, and shape public knowledge environments (Coeckelbergh 2025).

The anthropological significance of artificial intelligence, therefore, lies less in any claim to ontological equivalence and more in the redistribution of epistemic authority. This redistribution affects the conditions under which human agents exercise judgment. The human agent is no longer the primary locus of first-order epistemic engagement with phenomena but becomes, in many cases, a second-order evaluator of algorithmically generated outputs.

Within the Christian tradition, epistemic capacities are not incidental but constitutive of the human vocation. Concepts such as logos, synderesis, and conscientia have been used to describe the human orientation toward truth and moral discernment, particularly within the Thomistic tradition (Aquinas, Summa Theologiae; McFadyen 1990). If the exercise of these capacities is systematically mediated by computational systems, the question is not whether the human ceases to be human, but whether the conditions for their exercise are transformed in ways that are anthropologically and theologically significant.

3.3. The Anthropological Stakes of Delegation

The implications of epistemic automation can be illustrated across several domains.

In the domain of moral reasoning, artificial intelligence systems are increasingly used in contexts such as content moderation, risk assessment, and resource allocation. While human agents formally retain decision-making authority, the relevant information is often pre-structured by algorithmic processes. This introduces a shift from first-order moral judgment to second-order evaluation of algorithmic outputs. Such a shift has been recognised not only in philosophical debates but also in regulatory frameworks. The EU AI Act 2024 explicitly restricts certain forms of automated decision-making in morally significant domains, reflecting institutional recognition of the limits of delegating moral judgment to artificial systems.

In the domain of interpretive practice, large language models are capable of generating texts that exhibit high levels of fluency and contextual appropriateness. This has implications for disciplines such as theology, where interpretation is not merely instrumental but formative. As Vallor (2016) argues in the context of technomoral virtue, practices of interpretation contribute to the development of moral and intellectual character. When interpretive labour is partially or wholly delegated, the relationship between the agent and the practice is altered.

At the level of epistemic trust, artificial intelligence introduces a novel form of authority that is not grounded in traditional structures of accountability. Human epistemic communities rely on socially mediated trust relations, including expertise, institutional validation, and communal discernment (Fricker 2007). Artificial systems, by contrast, derive authority from performance characteristics such as accuracy and fluency, without corresponding mechanisms of responsibility or answerability. This shift has implications for theological accounts of knowledge, particularly those that emphasise communal discernment and tradition as conditions for reliable knowing (McFadyen 1990).

4. Fluency, Understanding, and the Deformation of Agency

4.1. Fluency Versus Understanding as an Anthropological Boundary

A defining feature of contemporary artificial intelligence systems is the capacity to generate outputs that are linguistically fluent and contextually appropriate without possessing understanding in any robust sense. This distinction has been widely discussed in philosophy of AI and cognitive science, where fluency is understood as the ability to produce coherent outputs, while understanding involves semantic grasp grounded in embodied and intentional processes (Coeckelbergh 2020).

The decoupling of fluency from understanding introduces a new epistemic condition. In human communication, fluency has typically functioned as a practical indicator of understanding, even if the two are not conceptually identical. Artificial intelligence disrupts this alignment by producing outputs that simulate understanding without the underlying cognitive processes. This distinction may be further clarified through a theology of the Word. Within the Christian tradition, logos refers not merely to linguistic output but to intelligible participation in truth (Augustine 1991). Fluency without understanding thus approximates what Augustine describes as a form of empty or idle speech, in which verbal production is detached from interior grasp. By contrast, intellectus denotes a participatory mode of knowing that involves transformation of the knower, often described in Thomistic terms as connatural knowledge (Aquinas 1947). Theologically, the risk introduced by AI is not simply the production of fluent language, but the redefinition of intelligibility itself, such that fluency becomes a sufficient proxy for understanding. In this way, artificial systems may mimetically reshape the epistemic norms by which human knowing is evaluated (Coeckelbergh 2020; Vallor 2016).

Within the Christian theological tradition, understanding is not reducible to cognitive performance. Augustine’s account of intellectus and Aquinas’s notion of connatural knowledge describe understanding as a participatory mode of knowing that involves transformation of the knower (Aquinas 1947; McFadyen 1990). Knowledge, in this sense, is not merely informational but formative.

The widespread availability of fluent outputs may alter the conditions under which understanding is cultivated. If epistemic success is measured primarily in terms of output quality rather than formative engagement, practices that sustain understanding may be displaced. Related concerns have been raised in broader critiques of technological mediation, which emphasise the potential erosion of practices that require sustained attention and embodied engagement (Vallor 2016; Coeckelbergh 2025).

4.2. Delegated Cognition as Agency Deformation

The concept of delegated cognition captures the transfer of cognitive tasks from human agents to external systems. While such delegation is historically ubiquitous, the delegation of higher-order epistemic capacities, including judgment, interpretation, and evaluation, represents a qualitative shift in the structure of human agency.

The notion of deformation is used here in a theological sense, referring to a structural alteration in the configuration of human agency. Christian theology has long recognised that agency can be formed or deformed through habituation and structural conditions (Waters 2006). The term deformation requires theological specification. Within the Christian tradition, the form of human agency is not morally neutral but teleologically ordered toward participation in truth and the good. To speak of deformation is therefore not merely to describe structural alteration, but to indicate a deviation from the proper ordering of intellect and constitutes the imago Dei (Aquinas 1947). This account presupposes a minimal theological anthropology in which human agency is oriented toward truth through the integration of intellectus and voluntas, sustained within conditions of finitude and dependence (McFadyen 1990). A transformation becomes a deformation when it systematically weakens these conditions, not by eliminating agency, but by reconfiguring its operative structure in ways that obscure its orientation toward truth, responsibility, and divine relation (Waters 2006). This deformation may also be understood in relation to the doctrine of the Fall, not as a series of discrete moral failures, but as a structural distortion of the conditions under which truth is apprehended and enacted. In this sense, epistemic automation does not introduce sin ex nihilo but may intensify patterns in which mediation obscures rather than orders the relation between the knower and the known (Aquinas 1947; McFadyen 1990). The present claim, then, is that epistemic automation introduces new forms of structural deformation by reshaping the conditions under which these capacities are exercised.

First, there is a deformation of attention. Epistemic practices that require sustained engagement with phenomena are increasingly mediated by representations produced by artificial systems. This may shift the locus of attention from the object to its algorithmic representation, a dynamic consistent with broader analyses of technological mediation (Coeckelbergh 2020; Vallor 2016).

Second, there is a deformation of responsibility. When epistemic processes are distributed across human and artificial agents, the attribution of responsibility becomes diffuse. This problem has been widely discussed in the ethics of AI, particularly in relation to accountability gaps in automated decision-making systems (Coeckelbergh 2020).

Third, there is a deformation of receptivity. The Christian tradition emphasises the importance of practices through which individuals become receptive to truth, including scriptural interpretation, communal discernment, and spiritual discipline. If these practices are systematically displaced by more efficient alternatives, the conditions for such receptivity may be weakened.

4.3. The Mimetic Dimension

The transformation of epistemic practice also has a mimetic dimension. Human cognition is shaped through processes of imitation, as emphasised in the work of René Girard (Girard 1965). Individuals learn not only what to know but how to know through engagement with models.

Girardian mimesis operates not only at the level of imitation but at the level of desire and rivalry (Girard 1965). The relevant question is therefore not whether human agents imitate the outputs of AI, but whether AI reshapes the structure of epistemic desire itself. If truth becomes increasingly mediated by systems that present themselves as neutral and authoritative, then traditional sites of epistemic conflict may be displaced rather than resolved. One possibility is that algorithmic systems function as epistemic scapegoats, repositories of judgment to which responsibility is transferred, thereby diffusing the tensions inherent in communal discernment (Girard 1986). In this case, the delegation of judgment to AI does not eliminate conflict, but redistributes it, weakening the communal practices through which truth is contested, negotiated, and collectively discerned (Fricker 2007). For theological anthropology, this shift is significant because formation is inherently mimetic. The development of judgment and discernment depends on participation in practices and communities that embody these capacities. The introduction of artificial models into this process alters the conditions under which such formation occurs.

5. Toward a Reconstructed Theological Anthropology

5.1. The Human Under Conditions of Epistemic Transformation

The preceding analysis suggests that theological anthropology must account not only for what the human is, but the conditions under which distinctively human capacities are exercised. This requires a shift from ontological comparison to what may be termed epistemic ecology, that is, the structured environment within which knowledge is produced, validated, and enacted. Such a shift is consistent with broader developments in philosophy of technology, which emphasise that technological systems reshape the conditions of agency rather than merely extending pre-existing capacities (Humphreys 2004; Coeckelbergh 2025).

Within this framework, the imago Dei is not denied or reduced but reframed in terms of its conditions of actualisation. The Christian tradition has consistently linked human distinctiveness to the exercise of epistemic and moral capacities in relation to truth and the good (McFadyen 1990; Waters 2006). However, these capacities are not self-sustaining. They require practices, institutions, and forms of life that enable their development and maintenance. Under conditions of epistemic automation, these supporting structures are subject to transformation.

This transformation calls for a more explicit theological account of epistemic formation. While moral and spiritual formation have been extensively developed within the tradition, including in virtue ethics and spiritual theology, the formation of epistemic capacities such as judgment, discernment, and interpretation has often been treated implicitly. Yet these capacities are central to the human vocation as understood within Christian theology. As Vallor (2016) argues in the context of technomoral virtue, practices that cultivate attention, reflection, and judgment are essential to sustaining ethical agency in technologically mediated environments.

A reconstructed theological anthropology must therefore articulate the conditions under which epistemic formation remains possible. This includes attention to educational practices, communal structures of discernment, and forms of engagement that resist the reduction of knowledge that is output. The challenge is not to reject epistemic automation but to situate it within a broader ecology that preserves the conditions for understanding as a formative process.

A further implication concerns epistemic vulnerability. Human knowing, within the Christian tradition, is characterised by finitude, fallibility, and dependence on others. These features are not deficiencies but constitutive aspects of creaturely existence (Fricker 2007; McFadyen 1990). Artificial intelligence systems, by contrast, are often designed to minimise uncertainty and maximise performance. While this has clear practical advantages, it may also contribute to the devaluation of epistemic humility and openness, which are essential to both philosophical inquiry and theological reflection.

The question, therefore, is not whether artificial systems can eliminate epistemic vulnerability, but whether their widespread use alters the normative expectations of knowledge in ways that make such vulnerability less intelligible or less valued. A theological anthropology attentive to epistemic transformation must address this tension directly.

This suggests the need for a more explicit account of creaturely knowing. Within the Christian tradition, human knowledge is constitutively marked by finitude, fallibility, and dependence on others, not as deficiencies, but as conditions that orient the knower toward grace and community (Fricker 2007; McFadyen 1990). Epistemic vulnerability is therefore not a limitation to be overcome, but a theological virtue that sustains openness to correction and transformation. The increasing normalization of AI-mediated certainty risks displacing these features by rendering dependence less visible and less intelligible. A theological anthropology adequate to the present condition must therefore defend epistemic vulnerability as a constitutive dimension of faithful knowing rather than as a deficit relative to computational efficiency (Vallor 2016).

5.2. Practical and Ecclesial Implications

In pastoral contexts, similar considerations apply. Consider the case of AI-assisted sermon preparation. A preacher who relies extensively on generated content may produce sermons that are rhetorically effective and structurally coherent. However, the process of preparation, traditionally understood as a site of theological and spiritual formation, is altered. The act of wrestling with scripture, engaging tradition, and discerning application is not merely instrumental but formative (Waters 2006). When this process is externalized, the preacher’s role shifts from participation to curation.

A comparable dynamic arises in contexts of pastoral discernment, including spiritual guidance, pastoral counselling, and confessional preparation, where reliance on pre-structured outputs risks substituting procedural adequacy for the relational and interpretive labor through which pastoral judgment is formed. In such cases, the question is not only the correctness of the output, but the formation of the agent that must bear responsibility for its use. AI-generated guidance may standardize responses while attenuating the interpretive labor through which pastoral judgment is cultivated (Vallor 2016). The use of artificial intelligence in sermon preparation, pastoral counselling resources, and liturgical planning thus introduces new forms of epistemic mediation. While such tools may enhance productivity, they also risk attenuating the practices through which pastoral agents are themselves formed. As Waters (2006) notes, theological reflection is not merely instrumental but participatory; it involves the transformation of the agent as well as the production of content.

The question is therefore not whether AI-assisted outputs are adequate in functional terms, but whether the processes through which they are generated contribute to or undermine the formation of agents capable of theological judgment. This distinction is central to any theological assessment of technology.

A viable epistemic ecology under conditions of AI mediation requires intentional structuring. At the communal level, practices of shared discernment must remain primary modes of judgment rather than being subordinated to algorithmic outputs (Fricker 2007). At the institutional level, theological education must preserve forms of assessment that privilege process over output, including dialogical reasoning and sustained interpretive engagement (Millgram 2015). At the technological level, artificial systems must be constrained to roles that support rather than replace first-order epistemic engagement, functioning as tools within practices rather than as mediators that redefine them (Coeckelbergh 2025). The aim is not to exclude AI, but to embed it within a hierarchy of epistemic authority that preserves the formative integrity of human agency. Such structuring may include requirements for human interpretive justification alongside AI-generated outputs, communal review processes for doctrinal or pastoral decisions, and pedagogical constraints that require first-order engagement prior to technological assistance.

5.3. A Non-Reactive Theology of Technology

A non-reactive theology of technology requires a shift from evaluating artificial intelligence as an external object to understanding it as a constitutive element within the structuring of human epistemic and social relations. As the preceding analysis has shown, epistemic automation does not merely redistribute cognitive tasks but alters the conditions under which judgment, responsibility, and authority are exercised (Coeckelbergh 2025; Humphreys 2004). The theological question is therefore not whether artificial systems possess agency, but how their integration reshapes the relational and normative structures within which human agency is formed.

Girard’s account of mimesis provides a crucial framework for analysing this transformation. Mimesis is not reducible to imitation at the level of behaviour but operates at the level of desire, structuring what agents take to be worth knowing, affirming, and pursuing (Girard 1965). In epistemic contexts, this extends to the formation of epistemic desire: the orientation toward truth, recognition, and authority within a community of knowers. What counts as credible knowledge and who is treated as a legitimate epistemic authority are not neutral facts but outcomes of socially mediated processes (Fricker 2007).

Artificial intelligence systems introduce a distinctive modification into this mimetic structure. They function as epistemic models that exhibit fluency, consistency, and apparent neutrality, yet they do so without participation in the reciprocal and contestable dynamics that ordinarily characterize human epistemic communities (Coeckelbergh 2020; Vallor 2016). Their outputs are often treated as authoritative, not because they are embedded in relations of accountability, but because they perform reliability according to technical criteria. This creates a form of authority that is structurally detached from the communal practices through which epistemic responsibility is ordinarily negotiated (Hauswald 2025).

In this context, Girard’s account of the scapegoat mechanism becomes analytically significant. The scapegoat functions as a site onto which communal tensions are projected, allowing conflict to be resolved through displacement rather than through direct engagement (Girard 1986). A structurally analogous dynamic can emerge in epistemic contexts shaped by AI. When judgment is delegated to algorithmic systems, responsibility for decisions may be displaced onto outputs that appear objective and non-contestable. This does not eliminate epistemic conflict but redistributes it, obscuring the role of human agents in its production and resolution.

Such displacement has important theological implications. First, it risks attenuating the formation of epistemic responsibility. Within the Christian tradition, practices of discernment are inherently communal and participatory, requiring agents to engage in processes of interpretation, disagreement, and mutual correction (McFadyen 1990; Waters 2006). If these processes are systematically mediated by systems that do not themselves participate in communal life, the formation of agents capable of bearing responsibility for truth may be weakened. Second, the displacement of conflict onto algorithmic systems risks obscuring the intrinsically dialogical character of truth-seeking. Knowledge, in this tradition, is not produced through the elimination of tension but through its ordered engagement within practices that sustain accountability and openness (Fricker 2007).

A non-reactive theology of technology must therefore move beyond both instrumental and adversarial models. Artificial intelligence is neither simply a neutral tool nor an external threat, but a structuring condition that reconfigures the epistemic ecology within which human agents operate. The task is not to eliminate such mediation, but to order it in ways that preserve the conditions for responsible judgment and communal discernment. This requires maintaining a hierarchy of epistemic authority in which artificial systems remain subordinate to practices that involve participation, accountability, and formation (Coeckelbergh 2025; Vallor 2016).

From this perspective, artificial intelligence functions as a diagnostic intensification of underlying anthropological dynamics. It renders visible the extent to which human knowing is always mediated, socially structured, and vulnerable to distortion. The theological task is therefore not to defend an unmediated conception of knowledge, but to articulate the conditions under which mediation supports rather than undermines the pursuit of truth. Such conditions include the preservation of communal practices of discernment, the cultivation of epistemic virtues, and the refusal to displace responsibility onto systems that cannot bear it (Waters 2006; McFadyen 1990).

6. Conclusions

This paper argues that the most productive engagement between artificial intelligence and theological anthropology lies not in ontological comparison but in epistemic analysis. The concept of epistemic automation identifies a pervasive transformation in the conditions under which human beings exercise the capacities that the Christian tradition has regarded as constitutive of the person: judgment, discernment, interpretation, moral reasoning, and receptivity to the divine. The distinction between epistemic authority and ontological status provides a framework for analysing this transformation without being drawn into speculative questions about machine consciousness. The distinction between fluency and understanding identifies an anthropological boundary that AI renders newly visible. The concept of agency deformation provides a mechanism for understanding how epistemic automation reshapes the human agent without destroying agency.

The constructive implication is that theological anthropology must move from reactive commentary on AI toward a general theory of the human under conditions of epistemic transformation. Such a theory would attend not only to what the human is but to what the human needs to exercise its distinctive capacities: conditions for epistemic formation, structures of communal discernment, practices that cultivate understanding rather than fluency, and environments that sustain the epistemic vulnerability through which genuine knowing occurs.

The scholars who will shape theological anthropology in the coming decades are not those who write about AI as a topic alongside others, but those who use AI as an occasion to reconstruct the architecture of the discipline itself. The task is not to defend the human against the machine, but to articulate the conditions under which the human can flourish as the kind of being that theological traditions, at their best, have always described: a being made for truth, formed through community, and oriented by grace toward a knowing that transforms.