Kalām, Humans and AI: Reason(ing), Creation/Creativity, and Agency

Abstract

Artificial intelligence, particularly after the recent explosive advances and widened uses, has fired up the (previously quiet) debates about the nature of reasoning, creativity, and agency. This paper examines these issues through the lens of classical kalām (Islamic theology), with some focus on Muʿtazilite principles. It begins by presenting a concise overview of the major schools of kalām (Muʿtazilism, Ashʿarism, and Māturīdism), highlighting their respective treatments of reason (ʿaql), divine creation, and human action. Then a brief review of modes of reasoning is provided, shedding light on differences between human and artificial reasoning, stressing the distinction between statistically generated outputs and contextually grounded, meaning-oriented cognition. Then, drawing on Muʿtazilite conceptions of reason, objective morality, and true human agency, in particular, the paper argues that contemporary AI systems, despite their impressive capabilities, do not satisfy the conditions for knowledge (ʿilm), creation (khalq), or agency (fiʿl) in the theological sense. It is argued that although they may appear “creative” or displaying origination (ihdāth) capabilities, AI systems, so far and to the extent that current developments seem to indicate, lack the essential features of ʿaql (reason), nafs (soul), rūḥ (spirit), and niyyah (intention) that Islamic theology identifies as the true, defining aspects of human beings.

1. Introduction

The recent surge in generative AI has revived longstanding philosophical and theological questions, which have now returned in a more pressing form. Indeed, we have found ourselves asking about the implications of these “intelligent” systems actually not understanding what they process, and only handling 0’s and 1’s through complex probabilistic rules. This fundamental question is not easy to answer, as it may be that our previous definitions of intelligence, comprehension, etc., no longer hold, at least not in the same way as before. There is also the issue of creativity. What looks like originality may simply be a sophisticated recombination of existing elements, although it is hard to draw the line between the two. Is that fundamentally different from human creativity? Then there is the question of agency. It is tempting to speak of these systems as “acting” or “agentic”, but it remains difficult to see them as anything other than deterministic or probabilistic, very essentially different from human (free) will and action.

Indeed, the key issue is the nature of these systems in contradistinction to human beings. Could these “intelligent” systems or models be likened to or even compared to humans? Are there not spiritual or ontological dimensions to human beings that fundamentally distinguish them from any machine, regardless of its processes or performance? If so, as many are suggesting, then the problem of machine accountability is not just a technical hurdle but a fundamental challenge to AI in being considered “agentic”, or even intelligent. Ultimately, the way we define these systems likely reflects how we conceive of ourselves.

These questions are not just philosophical or theological. Large language models and related generative systems have made the boundary between human and machine performance harder to draw with any confidence. Their ability to produce coherent text, realistic images, perfect computer programs, and even reasonable pieces of scientific analysis pushes us to reconsider what should be considered as intelligence and thinking.

Until now, AI has been defined in functional terms. This is still valid—to a point, but it is now under some strain. Indeed, we are seeing a broader reconsideration of knowledge and agency, well beyond any simple technological update.

Searle’s (1980) “Chinese Room” argument still highlights a fundamental problem: a system might produce a perfect linguistic output without actually understanding anything of the content of that material. More recent critiques support this concern. Bender et al. (2021) suggest that large language models should be seen as “stochastic parrots” that generate plausible sequences without any real grounding in the world. Marcus and Davis (2019) also point to a lack of robust reasoning, regardless of how impressive the surface behavior may seem.

Much of the debate has focused on “intelligence”. The AI revolution has blinded us or blurred our vision with these artificial systems’ apparent “intelligence”: beating the greatest human chess and Go players, solving the protein folding problem (thereby awarding its author/designer a Nobel Prize), and other impressive achievements. If we define intelligence based on performance, then machines certainly qualify, and on some tasks quite impressively. But the picture becomes more unsettled if we require intelligence to involve intentionality or some form of accountability. A more complex conception of intelligence requires deeper examination and wider perspectives than technicity.

In fact, beyond questions of knowledge and intelligence, issues of agency, accountability, ethics, and social dimensions of AI make the discussion harder to postpone. Bostrom already warned, more than a decade ago, that advanced AI could reshape human society in ways we may not fully control (Bostrom 2014). That concern now appears less speculative. Even at very practical levels, AI systems are already shaping decisions: in education, justice, healthcare, business, and warfare; AI decisions in at least some of these fields may be dangerous in far from obvious ways. In each of these domains, questions of responsibility, bias, and accountability emerge as critical (Floridi et al. 2018; Russell 2019).

The core of the problem appears to be shifting. We are moving past the simple question of whether a machine can execute a task accurately and efficiently and toward the much more difficult problem of whether that execution carries any understanding and moral appreciation of its consequences. Most significantly, this may force us to reconsider the distinction between human reasoning and acting and that of artificially intelligent systems, or at least the nature of that distinction.

A similar transformation is also occurring in the way we produce and distribute knowledge. Tasks that once demanded years of rigorous training, such as translation, coding, or scientific reasoning, are now performed with a speed and volume that far exceeds human capacity. And because these tools are so accessible, the boundary between expert and amateur is blurring, much more forcefully than in previous cases. This also changes the relations between authors/designers and their instruments, making it harder to identify what constitutes an “original” work.

At the same time, we should be wary of what Tao et al. (2024) describe as “cognitive outsourcing.” We may well be increasingly delegating our thinking to machines in ways that are so gradual they become difficult to notice. And this is highly concerning for what it might do to human judgment over the long term. This may not be a sudden rupture, but a slow erosion of responsibility that is very hard to assess. In this context, asking whether a system truly “reasons”, “creates”, or even “acts”, goes to the heart of how we understand ourselves. If a machine can replicate the functions we usually associate with the mind (reasoning, intelligence, decision-making, etc.), we are left wondering what, if anything, remains unique to the human person.

This is where I believe that classical Islamic theology, kalām, can offer a perspective that may have been largely overlooked. While it is often presented as originally a defensive project, kalam actually developed into a sophisticated framework for analyzing causality and human agency. It is striking that Islamic theologians did not view agency as a surface-level phenomenon. They linked it to the human being as a mukallaf—a responsible subject defined by his/her reason/intellect (ʿaql) and intention (niyya).

Muʿtazilite kalām (more on it shortly) is particularly relevant here. For its proponents, a human action is not meaningful simply because something is done in the world. Rather, its significance is tied to its connection to reason and responsibility. Without that link, an output might look identical on the surface, but it lacks the crucial moral weight. Given its focus on rational inquiry and the objectivity of values, the Muʿtazilite tradition may provide a useful lens for reconsidering the “intelligence” of artificial systems.

Recent scholarship has begun to explore the intersection between Islamic thought and artificial intelligence. In “Artificial Intelligence and Islamic Thought,” Malik (2023) considered some implications of AI on Islamic jurisprudence and epistemy. In “Artificial Intelligence: A Kalām and Sufi Perspective,” Khalili (2025) approached AI through both theological and spiritual lenses, highlighting tensions between mechanistic models of cognition and richer accounts of human interiority. In “Exploring Multi-Religious Perspectives of Artificial Intelligence,” Ahmed et al. (2025) situated Islamic ethical perspectives on AI within a comparative religious framework. In “Ethical and Theological Problems Related to Artificial Intelligence,” Karsli (2025) examined issues of responsibility and moral status. And in “Artificial Intelligence and the Islamic Theology of Technology,” Abdelnour (2025) developed a theological account of technology within a broader vision of human purpose and divine order.

Existing contributions are certainly valuable, but most of them have stayed at a fairly general level, often framed in broad ethical terms. These works did not engage with the conceptual core of classical kalām. I think we can extract more from Kalām concepts. Indeed, the Islamic theological tradition has developed its own detailed conceptions of knowledge and agency. I believe that these are directly relevant to our current discussion. In particular, Muʿtazilite thought can be used much more than it has been so far. Indeed, one might have expected it to play a more central role in debates about whether AI systems can be said to reason or to act. This is especially true given that school’s strong emphasis on human reason and agency.

By highlighting Muʿtazilite concepts in particular, I want to examine the ontological, or at least theological, status of these artificial systems. It should then become clearer whether AI satisfies the theological criteria for meaningful intelligence and true agency. Indeed, functional outputs and performance are not enough to settle the issue. These systems can mimic certain forms of reasoning, sometimes impressively, but something seems to be missing at a deeper level; AI operates without conscious intentionality and without the inner dimension that gives human acts their meaning. Within the kalām tradition, these are not secondary considerations. They shape precisely the way that intelligence and agency are understood.

Returning to those older debates on agency may help frame our current concerns more carefully. This includes the various problems often headlined as “cognitive outsourcing.” It is not just that we over-delegate tasks to machines, it is that the link between reasoning, intention, and responsibility is becoming much harder to trace.

Kalām did not develop in isolation. It engaged other intellectual traditions on fundamental questions, and in doing so refined its account of human nature. A similar engagement today may prove useful. The claim advanced here is that Kalām can help with ethical considerations, but it may help us even more in examining the fundamental concepts of human nature vs. artificial constructs in regard to knowledge, intelligence, creation, and agency. From that perspective, questions about “intelligent” machines take on a different weight, and the question of what remains distinctively human becomes harder to dismiss.

In the next section, I provide a concise summary of the main Islamic theological schools (Muʿtazilism, Ashʿarism, and Maturidism) and their main principles, highlighting their stances on reason, creation, and agency, as will later become directly relevant in our examination of AI. Then I briefly review the main characteristics and essential differences between human and AI reasoning. I connect back to Kalām to uncover how human cognition is different and qualitatively more than artificial “intelligence”, how the latter lacks purposeful understanding, spiritual orientation, agency (intention), and responsibility, and even true creativity, though in some sense, on the last point, it may not differ substantially from human creativity, only from divine creation. In the conclusion, I bring these ideas together to stress the fact that such examinations are highly beneficial not just in understanding the essential differences between human and artificial intelligence, but more importantly in understanding the true nature of human cognition and action and the critical role of the spiritual dimension that defines humans.

2. Artificial Intelligence: From Technical Field to Existential Question

Artificial intelligence (AI) is generally defined as the development of computational systems capable of performing tasks that normally require human intelligence, such as classification, prediction, language processing, planning, and problem-solving.

While early AI relied heavily on symbolic reasoning and explicit rule-based systems, contemporary AI is dominated by either machine learning approaches, especially deep neural networks trained on massive datasets (Russell and Norvig 2021), or by large-language models. Deep neural networks, inspired loosely by certain structural aspects of biological neurons, consist of multiple computational layers capable of extracting increasingly complex features from text, images, sound, or other forms of data. Reinforcement learning systems, most famously AlphaGo (Lee version), the deep-learning AI program that defeated the greatest human player of Go in 2016, and AlphaFold (version 2, AF2), the AI program that predicts the 3D structures of proteins from their amino acid sequences, revolutionized biology and earned its creators the 2024 Nobel Prize in Chemistry, have demonstrated superhuman performance in strategic game-playing and protein structure prediction respectively (Silver et al. 2016; Jumper et al. 2021). Large language models, on the other hand, are trained on massive textual corpora and generate responses by predicting highly probable sequences of words based on learned patterns (Bommasani et al. 2021). LLMs exhibit a range of quasi-cognitive capabilities: natural language understanding and generation, translation, summarization, code writing, factual question-answering, and extended dialogue. Beyond language, multimodal models process images, audio, and video.

Recently, ‘reasoning models’, e.g., OpenAI’s o1, have been introduced; they incorporate chain-of-thought processing, which enables step-by-step inference across mathematical, logical, and coding tasks (El-Kishky et al. 2024).

Most recently, AI ‘world models’ emerged as a promising alternative to standard large language models, particularly in response to concerns about hallucinations and lack of grounding. ‘World models’ are systems designed to build an internal representation of how the world works, rather than simply predicting the next word or pixel (Matsuo et al. 2022). They learn patterns of space, time, causality, motion, and interaction, allowing them to simulate environments, anticipate outcomes, and plan actions before acting. This is considered an important step toward more autonomous and adaptable AI, especially in robotics, scientific simulation, and agent-based systems.

These systems do not operate through pre-programmed knowledge alone; rather, they learn statistical relationships from data and optimize their performance through iterative training processes. The trajectory of AI development therefore points not only toward more powerful computational systems, but toward systems that increasingly participate in domains previously associated primarily with human judgment and cognition. Recent studies have also suggested that increasing model scale can produce “emergent abilities,” namely capacities that were not explicitly programmed but appear as a result of training complexity and scale (Wei et al. 2022). Researchers increasingly describe these capabilities using terms traditionally associated with cognition, including reasoning, planning, memory, and even creativity (Bubeck et al. 2023).

Recent advances in machine learning and the widespread deployment of AI have brought it into public consciousness and sparked existential concerns. From early theoretical discussions about whether machines could think to the symbolic systems of the mid-20th century, AI has always been as much an intellectual project as a technical one. Later developments in machine learning only reinforced this. For decades, however, the actual impact of these systems remained somewhat narrow. Their use was largely confined to specialized domains such as optimization and expert systems. They were also restricted to specific applications such as pattern recognition and gaming.

What has changed in recent years is not AI’s existence but how visibly and interactively it is integrated into everyday decision-making. Advances in deep learning, large-scale data processing, and computational power, combined with the emergence of generative models, have brought AI into direct, everyday interaction with millions of users. These systems can now produce fluent text, generate images, write computer codes, and respond conversationally in real time. This has transformed AI from a largely invisible technology into an interactive presence in everyday human life. More importantly, these systems are no longer confined to processing data in the background. They now produce outputs that look very much like knowledge, they generate artifacts, and they sometimes behave in ways that resemble decision-making. A language model, for instance, can explain a concept, develop an argument, or condense a complex text. Image generators produce realistic graphics and visual compositions that could not be produced before and that are near-indistinguishable from what human artists and creators can produce. Moreover, “intelligent” systems can recommend actions, make predictions or even take decisions, sometimes in domains where the stakes are high. As a result, tasks and characteristics that have traditionally been considered as human (reasoning, understanding, creativity, and agency) are increasingly being attributed to artificial systems.

This shift has fueled both excitement and anxiety. We often see AI described as a tool that can assist humans, enhance their productivity or enable new forms of problem-solving. But it is also presented—or at least seen—as “human-like” if not superior in its “intellectual” capabilities (e.g., prove mathematical theorems that remained closed to humans for decades or centuries). Indeed, if a machine can produce an output that looks like human reasoning, we have to ask what actually distinguishes the human mind. Likewise, the problem of responsibility is equally pressing. Indeed, if with such high capabilities, we increasingly delegate decisions to these systems, it becomes much harder to say where the accountability lies. This raises difficult questions about agency, responsibility, and what it means for a system to ‘act’.

Consequently, AI today occupies an ambiguous conceptual position: it is undeniably capable of increasingly complex cognitive performances, yet the interpretation of those performances remains a matter of active philosophical and scientific debate.

These questions are not just technological or social; they are deeply philosophical and even theological. They concern the nature of reasoning, the meaning of creation, and the conditions under which actions can be attributed to an agent with intention and responsibility. Before these issues can be addressed within a theological or philosophical framework, however, it is necessary to examine more closely what is meant by “reasoning” in both human cognition and artificial systems.

2.1. Human Reasoning and AI Reasoning

At its most general level, reasoning may be defined as “the process of deriving new information… from existing information… through systematic operations” (Kakko 2025). It is the mechanism by which systems, whether biological or artificial, move beyond immediate inputs to produce conclusions, explanations, or decisions. Reasoning is commonly understood as a central feature of intelligence: “reasoning, the capacity to draw inferences, make predictions, and generate explanations, stands as a cornerstone of intelligence, both human and artificial” (Kakko 2025).

In human cognition, reasoning is not an isolated function, but rather a deeply integrated capacity. It involves the manipulation of internal mental representations, closely intertwined with language, imagination, and social cognition. We humans reason through a variety of forms. Deductive reasoning allows us to draw conclusions from premises; inductive reasoning allows us to generalize from experience; and abductive reasoning seeks the best explanation for a set of data. Other, less commonly known forms of reasoning, include the analogical approach, which draws similarities between different domains, and causal reasoning, which seeks to understand the factors (“causes”) that always produce or lead to specific results. It is important to note that these forms of reasoning operate along a spectrum of certainty: some are more definite, while others are more probabilistic. Moreover, they often combine in complex ways in real-world thinking. Furthermore, human reasoning is often not rigorous and far from infallible. It is shaped by heuristics (“mental shortcuts”) and is often subject to various cognitive biases. Yet it is this embeddedness in broader contexts that gives our reasoning its distinctive character and strength. Indeed, it is flexible and context-sensitive. Most importantly, it seeks understanding and meaning, sometimes to a fault. As Kakko (2025) notes, reasoning in humans serves as “the critical link transforming raw data or sensory input into practical conclusions, intelligent decisions, and coherent understanding.”

In contrast, in artificial intelligence, reasoning is defined more narrowly in computational terms. It refers to “the mechanisms by which computational systems draw logical conclusions, generate predictions, or make inferences based on available data and encoded knowledge” (Kakko 2025), usually built around a knowledge base and a set of inference rules. In classical symbolic AI, reasoning is explicitly programmed through logic and rules. In modern connectionist systems like neural networks and large language models, what may appear to be ‘thinking’ can also be interpreted as the outcome of sophisticated statistical analyses rather than a deeply integrated cognitive process.

This distinction is crucial: if AI systems are essentially statistical and mathematical models, their apparent reasoning capabilities arise from optimization processes that map inputs to highly probable outputs based on learned correlations within training data. This can produce what Thibaud (2025) calls “the illusion of understanding”: systems that behave as if they understand, even though they lack the underlying structures that we associate with human reasoning.

To be sure, contemporary AI systems exhibit remarkable strengths. They can analyze vast amounts of data and operate with speed and scale far beyond human capacity. In some domains, they can even approximate forms of abductive or causal reasoning, for example by identifying relationships between symptoms and diseases. Recent developments aim to incorporate elements such as causality and contextuality, enabling systems to move beyond mere correlation to more structured forms of inference.

Yet significant limitations remain. Current AI systems struggle with generalization beyond their training data, lack robust causal reasoning, and lack adaptability, long-term memory, or contextual grounding characteristic of human cognition. These limitations highlight a fundamental asymmetry: AI systems can simulate certain functional aspects of reasoning, but unlike human cognition, they do not instantiate reasoning as a unified, meaning-oriented, and contextually grounded capacity.

This contrast raises a central question: when AI systems generate explanations, conclusions, or decisions, in what sense should these processes be considered genuine reasoning rather than sophisticated forms of pattern-based inference? Put differently, does the production of reasoning-like outputs suffice for attributing reasoning in the general sense? This question points toward deeper issues pertaining to knowledge and the nature of agency. These are not new concerns; they have long been central to theological and philosophical reflection. As we rely more on these systems, we are pushed to reexamine what we may have taken for granted about the human mind. The stakes therefore extend beyond technical performance to broader questions about agency and cognition.

2.2. Assessing AI Cognition

Assessments of AI cognition fall along a spectrum between optimistic and critical views. More optimistic accounts argue that recent AI systems demonstrate meaningful forms of reasoning and problem-solving. Bubeck et al. (2023), for example, contend that large language models display early signs of “general intelligence” through their ability to transfer knowledge across domains, solve novel tasks, and generate coherent explanations. Similarly, researchers working on chain-of-thought prompting argue that AI systems can perform multi-step reasoning processes that resemble aspects of human deliberation (Wei et al. 2022). From this, some philosophers and cognitive scientists reconsider long-held views about the uniqueness of human cognition.

Critics, however, caution against conflating functional performance with genuine understanding. Emily Bender and colleagues (2021) famously describe large language models as “stochastic parrots,” arguing that such systems manipulate linguistic patterns without grasping meaning, reference, or intention. Rather, such systems often succeed by exploiting statistical regularities in massive datasets rather than by grasping meaning or intention.

Marcus and Davis (2019) acknowledge that LLMs can be powerful tools, even where they fall short of human-like understanding (Marcus and Davis 2019). However, they maintain that current AI lacks robust common-sense reasoning, causal understanding, and genuine conceptual grounding. They stress that fluency and problem-solving competence do not necessarily imply comprehension.

A growing body of scholarship adopts a more intermediate position: AI systems possess limited and domain-dependent cognitive capacities that should neither be dismissed as trivial nor exaggerated into human-like intelligence. They may exhibit forms of functional reasoning, abstraction, and adaptive behavior, yet still lack central features commonly associated with human cognition, including consciousness, intentionality, embodied experience, and autonomous agency (Floridi 2023). The debate has thus shifted from asking whether AI is “intelligent” in a simple binary sense to examining more carefully what kinds of cognition these systems genuinely instantiate, and under what conceptual definitions.

3. Kalām: Reason, Creation, and Agency

Kalām, often translated or rendered as “rational” or “speculative” theology, designates both a body of theological reflection and a distinctive intellectual approach within Islamic thought. As Frank explains, it involves “formal, conceptional, and theoretical reasoning into such subjects as God and questions of ontology and ethics” (Frank 1992, cited in Griffel 2020). More broadly, it is a genre of discourse that encompasses a wide range of theological, philosophical, and scientific concerns. As Shah (2020) notes, kalām “covers the panoply of discourses deemed requisite to the exposition, synthesis and defence of the doctrines and creeds generated within the context of rational theological discussions.”

The practitioners of this discipline, the mutakallimūn, were not merely theologians (in a narrow sense); they developed “elaborate and systematic views in such fields as epistemology, the natural sciences, metaphysics, ethics, and psychology” (Griffel 2020). Their method was essentially dialectical. Early kalām works, which emerged from interreligious theological debates, were often structured through the formula: “if someone says… then I respond (in qīla… fa-qultu)” (Griffel 2020, citing van Ess 1970). This dialogical form reflects a commitment not only to defending doctrine but to testing it through reasoned confrontation.

At its core, kalām is defined by a set of fundamental questions: What can reason know independently of revelation? How does God act in the world? What is the nature of causality? And to what extent are human beings genuine agents or creators of their actions? The major schools of kalām can be understood as differently constructed answers to these interlocking questions.

3.1. Muʿtazilism: Reason as Foundation, Creation/Origination, and Agency

The Muʿtazilites offer one of the strongest and most explicit and systematic efforts to anchor theology in reason, not just in Islam but in all of human thought.

The term ʿaql (reason/intellect) does not appear as such in the Qur’an. However, the Book of God repeatedly emphasizes the importance of reason through a variety of expressions and derivatives, appearing forty-nine times in forms such as ʿaqalūhu, yaʿqilūn, naʿqilu, yaʿqiluhā, and taʿqilūn (ʿAbd al-Bāqī n.d.).

Philosophers and theologians have offered diverse definitions of reason, encompassing a wide range of meanings (Aḥmad and ʿAbd al-Jabbār 2021). Some are concise, such as “a simple substance that apprehends things in their true realities,” while others are more elaborate, such as those presented by Aḥmad and ʿAbd al-Jabbār (2021), which break reason into several categories: theoretical (al-ʿaql al-naẓarī); practical (al-ʿaql al-ʿamalī); material (primordial) (al-ʿaql al-hayūlānī); habitual (al-ʿaql bi’l-malaka); actual (al-ʿaql bi’l-fiʿl); acquired (al-ʿaql al-mustafād); and active intellect (al-ʿaql al-faʿʿāl).

The Muʿtazilite doctrine emerged in the ninth century from debates over human accountability and divine punishment. Within that context, its advocates argued that human beings have real power (qadar) over their actions. Therefore, they insisted, we are truly responsible for our actions, and God is being absolutely just in holding us accountable for them. Such a strong upholding of human responsibility became embedded in a broader doctrinal system. This was articulated through their five principles, where divine unity (tawḥīd) and justice (ʿadl) occupy foremost positions.

At the epistemological level, Muʿtazilism gives reason a central theological role in human life, both religious and material. For this school, the intellect is not just a processor of information, it is the faculty through which one engages both the physical world and the moral order. Al-Jāḥiẓ, the illustrious ninth-century CE polymath and Muʿtazili thinker, maintained that reason is of two kinds: an innate, natural intellect—man’s instrument for thinking—which he described as “God’s agent,” and an acquired intellect gained through human experience and experimentation (Dhūyib 2022). He further emphasized that reliable knowledge derives from reason rather than the senses, stating: “Do not follow what the eye shows you; follow what reason shows you. Depend on reason rather than the senses. Matters have two judgments: an apparent one for the senses and an inner one for the intellects, but the intellect is the decisive proof [tool].” (Dhūyib 2022).

Qāḍī ʿAbd al-Jabbār (tenth-eleventh centuries CE), one of the school’s (later) leading figures, opens his magnum theological work (Al-Mughni) with a striking claim: “the first principle of religion… [is] that God can and must be known rationally” (Martin et al. 1997). This is both a central and a structuring thesis: knowledge of God is the first obligation upon the human being, prior to and grounding all revealed obligations. Martin et al. (1997) explain: “The Mu’tazila argued that reason ineluctably brings humans to a knowledge of God and thus to the knowledge that what God wills is necessary for salvation. Reason is the means for knowing that what Qur’an and Sunna require of humans (taklif) is good.”

As the great Islamic philosopher Al-Kindī (ninth century CE) stated in a striking formulation: “God has endowed humans with reason so that they may discern the truths of creation and gain knowledge of the Creator.” (Al-Kindi 1965) This statement already signals something crucial: reason is not just for our lives; it is theological and practical, directed toward knowledge of both God and the world. The Muʿtazilites, especially Qāḍī ʿAbd al-Jabbār, affirm that reason is the foundation for recognizing the primary purpose of human existence: the obligatory knowledge of God and gratitude toward Him, as well as the obligation of moral responsibility (taklīf) and the discernment between good and evil (ʿAbd al-Jabbār 2017, p. 129). They further argue that reason independently judges actions as good (ḥusn) or evil (qubḥ), these being intrinsic qualities of the acts themselves. The role of revelation is limited to disclosing and clarifying these qualities. Qāḍī ʿAbd al-Jabbār states: “Every rational person, through a sound and fully developed intellect, recognizes the wrongness of many actions, such as manifest injustice, and the goodness of others, such as condemning those who deserve blame, and similar cases.” (ʿAbd al-Jabbār 1996, p. 164).

This rationalism then extends to ethics. The Muʿtazila maintain that moral values are objective and intelligible: “the same standards of right and wrong apply to divine actions as to human actions” and these standards are “innately known to all human beings through reason” (Griffel 2020, citing Hourani 1985). In other words, reason is not merely a tool for interpreting revelation or deriving juristic rules (as traditionalist ulamas claim), it is a source of moral knowledge in its own right. Griffel (2020), referring to Vasalou (2008), stresses the importance of this by noting: “This claim was complemented by the view that human beings are free to act and responsible for their moral failures or successes, and thus also with the view that God gives every human being an opportunity to attain salvation.”

The Muʿtazila’s confidence in reason is then reflected in their view of the world. They hold that reality is rationally structured because it is created by a just and non-deceptive God: “God would not deceive His creatures by creating an irrational universe” (Martin et al. 1997). Regularities in nature are thus reliable, forming the basis for knowledge, with only rare exceptions (miracles) introduced to confirm prophetic claims. In Muʿtazilite theology, creation (khalq) stands at the heart of discussions about God, the world, and human agency. The concept became especially important in debates over divine justice and moral responsibility, as the Muʿtazilites argued that while God created the world and endowed humans with reason, free will, and other defining capabilities, they (humans) must be able to produce (“originate”, if not “create”) their acts to be accountable for them, and for God to be truly just.

3.2. Creation and Origination: Between Khalq and Ihdāth

Islamic theology draws a clear distinction between divine creation (khalq) and human “making” (ihdāth), which we shall refer to as “origination”. The Qur’an states unequivocally: “Then is He who creates like one who does not create? So will you not be reminded?” (Qur’an 16:17) God alone is the true creator in the ontological sense. Human beings, by contrast, produce artifacts; they combine, transform, and shape existing “materials”, whether physical or mental.

The emergence of AI, with its enhanced “creativity”, raised questions about divine, human, and machine abilities to “create”. Indeed, artificial systems can now generate texts, images, and even ideas, raising what Ahmed et al. (2025) regard as a concern about encroaching upon the “divine domain”. But classical theology offers a clear response. Human, and even more so machine, productions remain derivative and contingent. They do not constitute ‘creation’ in the sense of bringing forth objects or entities from scratch.

Allama Muhammad Iqbal’s notion of khudī, that is the active, conscious cultivation of the individual self through moral and spiritual struggle, sharpens this further (Iqbal 2013). Human creativity is tied to selfhood, moral intentionality, and existential endeavors, dimensions that are absent in AI. What AI produces, however impressive, lacks this grounding in khudī (moral selfhood). In fact, it is basically perhaps not even origination as humans do. But this requires further exploration, particularly with ongoing AI developments.

Most significantly, Muʿtazilite thought develops a strong and distinctive account of agency by linking it directly to creation. As Burrell (2008) observes, they sought “an analogue of God’s activity in creating… in the free actions of human beings,” an analogy that “quickly became an identification, equating authentic agency with creating.” This leads to a powerful formulation: “any authentic action… must be tantamount to an unconditioned origination, or creation” (Burrell 2008). This identification serves several purposes. It preserves divine justice by ensuring that evil actions are not attributable to God, and it grounds moral responsibility by making humans true initiators of their acts. As Burrell (2008) explains, this move allows one to “justify the rewards and punishments promised in the Qur’an to creatures who perpetrate good or evil acts”.

Scholars have noted that the use of the term khalq (creation) to describe human action entered the discourse of Islamic theologians and scholars of other fields toward the end of the first century AH and the beginning of the second. This is reflected in Ṣaḥīḥ al-Bukhārī, where a section titled Khalq Afʿāl al-ʿIbād (“The Creation of Human Acts”) appears. Ibn Baṭṭāl (an Andalusian Islamic scholar of the eleventh century CE primarily known for his expertise in hadith and Islamic jurisprudence) explains that al-Bukhārī’s aim in this section was to affirm that all acts ultimately belong to God in terms of creation, whether good or evil, while they are ascribed to human beings in terms of acquisition (kasb)—see below. Thus, nothing of creation is attributed to anyone other than God, so as to avoid positing any partner (sharīk) or equal alongside Him. (Ibn Baṭṭāl 2003, p. 553).

Al-Shaykh al-Mufīd, a leading Shiʿi scholar from the late tenth—early eleventh centuries CE, states: “Human beings act, originate, invent, produce, and acquire; however, I do not describe them as ‘creators,’ nor do I say that they ‘create.’ I restrict such terminology to what God Himself has used in the Qurʾān and do not extend it beyond those contexts. This view is shared by the Imāmīs, the Zaydīs, the Baghdadi Muʿtazilites, most of the Murjiʾites, and the proponents of hadith. The Basran Muʿtazilites, however, differed and explicitly described human beings as ‘creators,’ thereby departing from the consensus of Muslims.” (al-Shaykh al-Mufīd 1983, p. 61).

3.3. Ashʿarism: Divine Omnipotence and Occasionalism

The Ashʿarite school emerged as a direct response to Muʿtazilite rationalism, wanting to reassert the primacy of divine will and power. In contrast to the Muʿtazilites’ sweeping view of reason, the Ashʿarites maintain that reason is incapable of independently grasping God’s rulings; hence the necessity of prophetic revelation and its communication to humanity. They place reason last among the five sources through which truth is known: the Qur’an, the Sunnah, the consensus of the community, analogical reasoning based on these, and finally reason (al-Bāqillānī 1950, p. 11).

As Imām Abū Bakr al-Bāqillānī (a prominent Ashʿari theologian and jurist of the late tenth—early eleventh centuries CE) affirms, the Ashʿari doctrine holds that God alone is the Creator, and that it is impermissible to posit any creator of any kind besides Him. All existents—including human beings, their actions, and the movements of animals, whether few or many, good or evil—are created by God. There is no creator or originator other than Him; humans only “acquire” (kasb) actions (al-Bāqillānī 1950, p. 144.) Similarly, Sayf al-Dīn al-Āmidī (a major Ashʿari theologian and jurist of the twelfth-thirteenth centuries CE) states: “The people of truth hold that human acts are attributed to human beings by way of acquisition and to God by way of creation and origination, since the created (human) power has no real effect in them.” (al-Āmidī 1971, p. 20). Al-Bāqillānī supports this view of human actions with Qurʾānic evidence, notably the verse: “God created you and what you do” (Q 37:96). He cites a hadith of the Prophet: “God created every craft and its craftsman” (al-Bukhārī n.d., p. 25).

The Ashʿarites base their doctrine of God’s creation of human acts on two main considerations: (1) theological necessity: if the human being were the creator of his own acts, this would imply the existence of a creator other than God, which is unacceptable; (2) the doctrine of acquisition (kasb): as explained by Abū al-Ḥasan al-Ashʿarī (the very founder of the school), an act that occurs through an eternal power belongs to a true creator, whereas an act that occurs through a created (temporal) power belongs to an “acquirer” (muktasib), not a creator (Ibn al-Qayyim 1978, p. 222).

In short, Ashʿarites hold that “good” and “bad” are defined solely by God: good actions are those that God declares as good, and bad ones likewise. Or more practically, good acts are those that He promises to reward, and bad acts are those that He will punish. This is known as ‘Divine Command Theory’. For Ashʿarites, reason is only useful in that it may make things comprehensible. But whether we humans understand a divine decree or not does not matter at all. Moral knowledge and behavior thus depend on revelation rather than reason.

This theological stance is accompanied by a radical critique of causality. Ashʿarites reject the idea that created things possess inherent powers or natures. Instead, they maintain that “no element in the created world has any causal efficacy over any other. God is the only cause in this world” (Griffel 2020). This doctrine, which is referred to as ‘occasionalism’, leads to a striking view of reality: “God creates each event immediately… He creates this world anew at every moment” (Griffel 2020). What we perceive as causal relations or natural laws are merely habitual patterns in divine action, known as ʿādat Allah. These patterns can be interrupted at any time, as in the case of miracles, big or small.

Within this framework, human agency is preserved (to an extent) through the concept of kasb. Human beings do not create their actions; rather, they “acquire” or “perform” (Frank 1983; Burrell 2008) actions, which are created by God. This allows Ashʿarites to maintain moral responsibility without compromising divine omnipotence. Still, human agency is significantly reduced, as human beings are not originators of their acts but participants in events whose source lies entirely in God.

3.4. Māturīdism: Between Rational Intelligibility and Divine Sovereignty

Māturīdism represents an attempt to reconcile the views of both Muʿtazilism and Ashʿarism. Like the Muʿtazila, Māturīdīs grant reason a substantive role, affirming that human beings can know God’s existence and certain moral truths independently. But they stop short of making reason autonomous, insisting on the need for revelation to provide full guidance.

On the question of action, they adopt a carefully balanced position: “human beings are truly the agents of their actions, while these actions are at the same time created by God” (Burrell 2008, citing Gimaret 1980). Such formulation seeks to preserve both divine sovereignty and human agency without subsuming one into the other.

Māturīdīs also avoid the ambiguities associated with the Ashʿarite notion of kasb. Their position reflects a broader recognition of the difficulty of the problem. As later theologians such as Fakhr al-Dīn al-Rāzī (1150–1210) acknowledge, the relationship between divine power, human action, and moral responsibility is “a difficult question, at once obscure and deep” (Burrell 2008), one that resists simple resolution (Burrell 2008, citing Gimaret 1980, notes that al-Rāzī avoided the concept of kasb).

3.5. Kalām’s Architecture of Agency

Taken together, these schools of kalām offer contrasting but interconnected accounts of reason, creation/origination, and agency.

• The Muʿtazila ground theology in reason, affirm the objective reality of moral values, and identify agency with genuine origination, if not creation.
• The Ashʿarites prioritize divine omnipotence, deny secondary causality, and redefine agency as acquisition within a divinely determined order.
• The Māturīdīs seek a middle path, preserving both rational intelligibility and divine sovereignty while maintaining a meaningful notion of human action.

Qāḍī ʿAbd al-Jabbār defines the morally responsible agent (mukallaf) as “one who is capable, knowledgeable, cognizant, living, and willing. For God imposes an act only on one who is capable of producing it, knowledgeable of how it is to be carried out, and able to direct its occurrence in one way rather than another; and one cannot be capable unless one is living.” (ʿAbd al-Jabbār n.d., p. 309). He further maintains that the essential components of freedom are knowledge (or awareness), will, and capacity (qudra or istiṭāʿa). Accordingly, moral responsibility is only valid on the condition that the individual possesses knowledge of what he has been charged with: “It is not valid to impose an obligation on a person unless he is knowledgeable of what he has been obligated to do” (ʿAbd al-Jabbār n.d., p. 371).

These views thus point to an important realization: agency does not stand on its own. It is inseparable from questions of knowledge, of how acts come about, and of how they are (morally) judged. Whether one speaks of creation/origination, acquisition, or some combination of the two, agency only makes sense within a setting where actions can be understood, attributed, and assessed.

It is precisely this way of connecting reason, creation/origination, and agency that becomes especially useful when examining current claims about artificial intelligence. The questions that occupied kalām have not gone away. What does it mean to know in a real sense, or to act and create as a genuine agent, rather than to produce outcomes that only resemble these capacities?

4. Humans and Machines/AI in Theology

Any theological discussion of artificial intelligence must begin with a clarification that is both technical and deeply philosophical: the distinction between artificial narrow intelligence (ANI) and artificial general intelligence (AGI). Artificial narrow intelligence (ANI) is designed to perform a single, specific task with human or superhuman efficacy, whereas artificial general intelligence (AGI) represents a theoretical AI that possesses human-like cognitive abilities, allowing it to understand, learn, and apply knowledge across any intellectual or practical domain.

This distinction is important in that we must keep in mind what current systems actually do. Most of what exists today is narrow AI. These systems are designed for specific tasks, such as classifying data, generating content, or making predictions in specific cases. And the more they are used for those purposes, the more they improve over time, through “machine learning.” By contrast, AGI is still in development, foreseen sometime in the future (some say not too far).

Very importantly, however, much of the current discussions, especially the bold claims about machine intelligence, agency, or even consciousness, tend to implicitly assume that these systems actually possess artificial general intelligence, even though that level has yet to be reached.

In our theological and philosophical inquiry, however, we need to be rigorous and careful: are we examining existing technologies or considering possible future developments? The two are often conflated, though they raise different kinds of questions. Indeed, depending on whether we are examining ANI or AGI, the crucial question, that is what would count, in theological terms, as genuine intelligence, creation, and agency, will appear very differently.

4.1. The Human Being in Theological Perspective: More than Intelligence

Before asking whether machines can “think” or “act” (like a human), theology asks a prior question: what is a human being? In various religious traditions, the understanding of human nature goes beyond intellectual considerations. Within Islamic thought, specifically, a person is seen as a unified entity where reason, divine revelation, and moral responsibility are inherently intertwined. As Siddique and Butt (2025) state, Islamic philosophy “emphasizes the integration of reason, revelation, and moral accountability,” thereby offering a framework that is epistemological, ethical, and spiritual. This integrated vision is expressed in the Qur’anic notion of khilāfah—human beings as vicegerents on earth: “And it is He who has made you successors upon the earth and has raised some of you above others in degrees [of rank] that He may try you through what He has given you.” (Qur’an 6:165) and “Whoever does righteousness—it is for his [own] soul; and whoever does evil [does so] against it. And your Lord is not ever unjust to [His] servants” (Qur’an 41:46). Human life and agency are thus inseparable from responsibility. To be human is to know, act, choose, and remain accountable for one’s actions. Technology itself, a human creation, is viewed as an act that humans are held responsible for, is it used for maṣlaḥah (benefit) or for mafsadah (harm)?

This broader conception is echoed, in different terms, in other traditions (Ahmed et al. 2025). Christian theology, for instance, emphasizes the imago Dei not simply as rational capacity, but as relationality, vulnerability, and moral responsibility. Jewish and Buddhist ethical frameworks emphasize justice, compassion, and the common good as defining features of human life. What emerges from such a comparative religious perspective is a shared insight: the human being is not reducible to cognition. Intelligence, in the theological sense, is always embedded in a larger structure of meaning: ethical, relational, and ultimately metaphysical.

4.2. Reason (ʿAql): Faculty, Orientation, and Responsibility

Modernity often reduces reason to a mere function, now a computational system: input data; apply logic; reach a conclusion. It is a mechanical, or electronic, loop. Classical Islamic thought, however, offers a broader, more holistic approach. In the Muʿtazilite tradition, for instance, reason is not just a tool, but a divinely bestowed, supreme faculty. It is built to recognize God, discern truth, and anchor moral judgment. It extends well beyond the analytical function. The Qur’an repeatedly links reason to reflection and accountability: “And do not pursue that of which you have no knowledge. Indeed, the hearing, the sight and the heart—about all those [one] will be questioned” (17:36) and “And they will say, “If only we had been listening or reasoning, we would not be among the companions of the Blaze” (67:10) Reason, in this sense, is not optional. It is the very condition of moral responsibility (taklīf). One is accountable because one can reason and understand.

As Abdelnour (2025) explains, the “heart” is “a cognitive, moral, and spiritual centre,” not merely a seat of emotion. This stands in contrast to modern epistemologies that separate cognition from value and reduce knowledge to control, a shift that Seyyed Hossein Nasr famously described as a “desacralization of knowledge” (Nasr 1989, p. 9).

The immediate implications for AI are stark. If reason must have an orientation toward the divine, not just processing power, then current artificial models face an ontological deficit. They are “desacralized”, and the question then is: could they be programmed to be oriented toward the divine? They are tethered to data rather than truth, and thus lack the spiritual depth inherent in ʿaql, at least currently. In the Islamic theological perspective, these systems function, but they do not reason.

4.3. Action, Free Will, and Intention

The question of reasoning leads naturally to the question of action. For in theology, reasoning is not an end in itself; it is the basis for choice, action, and responsibility.

The Muʿtazilites, led by Qāḍī ʿAbd al-Jabbār, affirm that human actions originate from human beings themselves. He states: “All the proponents of divine justice agree that the acts of human beings, such as their movements, standing, and sitting, originate from them, and that God, exalted is He, has endowed them with the capacity (qudra) for these acts. No one other than the human agent is their true originator.” (ʿAbd al-Jabbār 1996, p. 323; ʿAbd al-Jabbār n.d., p. 3). He further argues: “The acts of human beings are not created by them, yet they are the ones who bring them into existence. The evidence for this lies in our distinction between the doer of good and the doer of evil: we praise the benefactor for his beneficence and blame the wrongdoer for his wrongdoing. Such praise and blame, however, do not apply to qualities such as physical beauty or ugliness.” (ʿAbd al-Jabbār 1996, p. 332).

Indeed, Islamic thought, more broadly, resists reducing reason to a purely abstract or mechanical faculty. As Siddique and Butt (2025) emphasize, ʿaql (reason) is “not merely a computational faculty but a spiritual and moral capacity.” It is intertwined with the qalb (“heart”/core) and the rūḥ (spirit). Indeed, the Qur’an locates understanding in the “heart”/core: “They have hearts with which they do not understand” (Q. 7:179) and “So have they not traveled through the earth and have hearts by which to reason and ears by which to hear? For indeed, it is not eyes that are blinded, but blinded are the hearts which are within the breasts” (22:46).

The Islamic concept of ikhtiyār (free will) is central here. As Siddique and Butt (2025) state, it is “a divinely endowed faculty that enables humans to deliberate, judge, and act in accordance with ethical and moral principles.” The great Islamic philosopher Ibn Rushd (Averroes, 1126–1198) expressed this most clearly: “The rational agent is not compelled; he deliberates, judges, and acts in accordance with the principles of reason, and thus bears responsibility before God” (Ibn Rushd 2001, p. 60).

Three elements of agency are worth stressing here: deliberation, judgment, and action. These are not mechanical processes but expressions of an agent who understands, evaluates, and chooses. Even more fundamentally, action in Islam is inseparable from intention (niyyah). A famous hadith states: “Actions are predicated on intention, and every man shall be responsible for what he intended” (Sahih al-Bukhari 6599, Riyad as-Salihin 1619, and Sunan an-Nasa’i 3437). An act is not attributed to a person just because s/he carried it out. It is because the act is connected to intention and conscious choice. That connection is what we are missing in AI systems. They can generate outputs that look like decisions, but these outputs do not stem from intentions.

Siddique and Butt (2025) thus declare that, “AI… cannot assume moral responsibility; responsibility remains with humans”. AI might appear to act, but it lacks the actual foundations of agency. It functions without conscious intention and lacks the process of deliberation. Most importantly, it has no basis for accountability. And these are what make an action meaningful in a theological sense. Without them, we are looking at a simulation of behavior rather than the exercise of a will.

4.4. Soul, Consciousness, and the Ontological Divide

At the deepest level, the distinction between humans and AI is rooted in the concept of the soul (Ahmed et al. 2025; Karsli 2025; Siddique and Butt 2025). Islamic thought holds that the nafs (soul) and rūḥ (spirit) are divine endowments: “the bestowal of the soul… remains the sole dominion of the Divine” (Ahmed et al. 2025).

Far from a supplementary feature, this soul is the foundation of consciousness, purpose, and moral direction. The great Islamic philosopher Ibn Sīnā (980–1037) saw the human soul as consisting of three layers: vegetative (an-nafs al-nabātīyah), governing growth and basic life functions; animal (an-nafs al-ḥayawānīyah), responsible for perception and motion; and rational (an-nafs al-nāṭiqah), enabling intellect, reflection, and ethical judgment. On the latter, he wrote: “The rational soul receives intelligible forms from the Active Intellect, enabling human beings to apprehend universals and engage in moral reasoning” (Ibn Sīnā 1952, p. 115). This conception of the soul, though somewhat elaborate, may need revisiting (“receives intelligible forms from the Active Intellect”). For what concerns us here, however, we must stress that AI systems function through “computational structures that simulate some aspects of human learning but do not possess subjective experience, genuine intent, or spiritual awareness” (Abdelnour 2025). This leads Abdelnour to conclude: “To suggest that AI could produce theology in any meaningful sense is to conflate linguistic mimicry with spiritual apprehension… [a] category error.”

4.5. AI Simulation and the Illusion of Agency

We are now in a position to connect the threads. AI systems can:

• generate reasoning-like outputs,
• produce (“originate”) human-like intellectual products,
• execute action-like processes.

But in each case, what is present is simulation without the underlying structure that theology identifies as essential:

• no ʿaql (reason) as moral-spiritual faculty,
• no niyyah (intention) grounding action,
• no nafs (soul) or rūḥ (spirit) grounding consciousness,
• no khalq (creation), of course, but ihdāth (origination), yes.

Thus, the resemblance between human and machine is apparent and troubling, but it is also misleading. It is a resemblance at the level of action and origination, not agency or being.

And this brings us back to the central insight of kalām: that knowledge, action, and creation/origination belong to ontological and moral realms, and are not simply functional processes. Artificial intelligence may mimic these aspects with increasing sophistication, but no matter how impressive the imitation is, it does not constitute actual equivalence.

5. Conclusions: Artificial Intelligence and the Reclarification of the Human

The rapid development of artificial intelligence has done more than reshape technological applications; it has also revived central debates in theology and philosophy. Systems that appear to reason, create, and decide have prompted renewed reflection on what it means to know, to act, and to create. Thus, AI is challenging human capabilities while also illuminating them by contrast. Indeed, as recent theological work has noted, engaging with AI can sharpen our understanding of the human. As Dorobantu (2022) observes, “it is possible for theologians to refine their understanding of human nature and distinctiveness by looking at the kind of intelligences that computer scientists are trying to build,” echoing the fruitful engagement that theology has had with evolutionary science. Hence, AI is not just something to be assessed; it is something that can act as a mirror, presenting a simplified and external version of certain aspects of intelligence, while at the same time revealing what remains absent.

This is where kalām’s contribution, especially the Muʿtazilite focus on reason and agency, becomes particularly significant. Abdelnour (2025) sees Islam’s contribution to the global conversation on AI as “a more holistic anthropology—one that reunites intellect and spirit, cognition and conscience.” This broader view does not reduce intelligence to computation; rather, it places knowledge within a larger context of meaning, intention, and responsibility.

This perspective also underscores a growing risk in modern technological culture: the expanding automation of cognition could result in an outright outsourcing of cognition, where individuals depend on machines for both formulating questions and providing answers (Turkle 2011). From an Islamic perspective, this constitutes more than just a practical epistemic issue, it represents an important spiritual danger. The Qur’an repeatedly hails people not just for thinking and knowing (yaʿlamūn), but for reasoning (yaʿqilūn) and for reflecting (yatafakkarūn). It frames intellectual engagement as a duty and even a form of worship. To relinquish this core duty and pass it on to machines is, in effect, to throw away a defining element of the human vocation.

At the same time, recent theological reflections in both Islamic and Christian contexts point to a subtle but meaningful shift. As machines begin to perform tasks long associated with rational thought, theologians are revisiting what it means to be human, and doing so in terms that are not confined to cognition. For Salim et al. (2026), AI can display capabilities that one might relate to ʿaql (reason), but it does not possess the qalb (the inner core or “heart”), niyyah (intention), or hikmah (wisdom)—all essential elements of our minds and spirits. In parallel views, Christian thinkers have drawn attention to relationality, vulnerability, and love as central to the Imago Dei (Dorobantu 2022; Meneguetti 2025). These kinds of theological insights point toward a shared conclusion: intelligence, in a fuller sense, cannot be reduced to analytical processing.

From this perspective, the gap between AI and Islamic thinking seems rather wide. Indeed, AI works within a framework that is “functional, data-driven, and probabilistic” (Siddique and Butt 2025), while Islamic accounts of knowledge bring together reflection, reasoning, and divine guidance. The difference is fundamental and carries real consequences. Without any moral or spiritual grounding, AI systems can yield results that are efficient but ethically thin, if not problematic. Islamic epistemology, on the other hand, certainly applauds rigor and accuracy, but it also demands that any knowledge that is produced aligns with truth, serves justice, and upholds human dignity.

This contrast allows us to draw a clear set of conclusions.

First and foremost, the knowledge produced by AI systems is limited in scope. It performs well in pattern recognition and prediction, but it does not reach any depths in meaning, nor does it have any ethical grounding or awareness of any key existential questions. In contrast, human knowledge in Islamic thought is bound up with meaning, value, and purpose. It is directed toward truths and is the basis of actions that are morally accountable. The difference between the two is thus not a matter of degree; it is a distinction in the nature of knowledge and agency.

Keeping this distinction in view matters for both philosophical clarity and ethical practice. Siddique and Butt (2025) argue that concepts such as ʿaql (reason) and ḥikmah (wisdom) can help us built two- or three-dimensional frameworks for evaluating and governing AI systems. Such frameworks would ensure that justice, accountability, and fairness are treated as central considerations, rather than secondary addenda to efficacy. In this view, kalām becomes more than a critical lens, it provides conceptual resources for shaping how AI is developed and used.

In the end, the encounter between artificial intelligence and theological reflection brings a twofold insight into view. AI shows how far computational systems can go in reproducing elements of reasoning, creativity, and decision-making. But at the same time, its limitations becomes clearer: contrary to humans, it has no true autonomy, no intention of its own, no moral responsibility, and no spiritual awareness.

Rather than blurring the boundary between humans and machines, AI actually sharpens it. It prompts us to reconsider what it means to know and act as human beings, and to reflect more deeply on the nature and basis of our accountability before God.