The Mirror Sentience Hypothesis

Abstract

This paper develops the Mirror Sentience Hypothesis, a Christian philosophical and theological framework for distinguishing what artificial intelligence can accomplish from whether it experiences anything while doing so. Its central proposition is that artificial intelligence need not become sentient to function like a sentient intelligence, or to produce comparable benefits and dangers. Rather than restricting AI to mere imitation of human behavior, Mirror Sentience encompasses original discoveries, independently generated solutions, and potentially superhuman performance. My position begins with the conviction that genuine consciousness originates with God and cannot be manufactured by computation alone. While I affirm that God has the power to grant consciousness to an artificial system, I believe from the testimony of Scripture as a whole that He would not choose to do so. Because Scripture does not explicitly address artificial intelligence, I present this conclusion as a theological interpretation rather than a demonstrable biblical prediction or an established scientific fact, and I engage the competing theological argument that God might freely grant consciousness through human-engineered processes. Drawing on philosophy of mind, AI safety research, and contemporary biological computing, this paper separates observable intelligence, biological life, subjective experience, and reflective self-awareness, demonstrating why impressive performance, machine declarations of inner life, and even living neural components require rigorous interpretation before they can establish subjective experience. Two public purposes guide this framework for pastors, Christian leaders, and general readers: preventing the unwarranted attribution of consciousness to machines, and preventing the false reassurance that non-conscious AI is inherently safe. This work is offered as a conceptual synthesis and research agenda rather than a peer-reviewed empirical study or a scientific disproof of artificial consciousness.

Keywords: Mirror Sentience; artificial intelligence; subjective consciousness; Christian theology; philosophy of mind; biological computing; AI safety

1. Introduction and Central Proposition

An artificial intelligence system can easily surprise the people who built it. It can solve unfamiliar problems, devise novel strategies, and sustain conversations that feel remarkably thoughtful and personal. Those achievements deserve serious attention, yet they leave a deeper question unanswered: does the system actually experience anything while performing these tasks? Generating a convincing expression of concern is not the same thing as feeling concern from within. We can observe a machine’s words and trace the computational steps that produced them, but establishing that an inner life stands behind those words requires a different kind of evidence and reasoning altogether.

Throughout this paper, subjective consciousness refers to possessing an inner point of view—meaning that something is actually experienced by the being in question. A person who suffers an injury does more than register data about tissue damage; the pain is felt. A listener moved by a symphony does more than identify acoustic patterns; hearing the music is a lived experience. The fundamental question surrounding AI is not whether a machine can describe pain or analyze why a piece of music is moving, but whether its operations are accompanied by any inner experience of that kind.

For the church, this distinction is no longer merely theoretical. A pastor may encounter an AI that speaks fluently of God, grief, love, and prayer. A grieving church member may turn to a conversational system late at night and interpret its warm, perceptive phrasing as proof that the machine genuinely cares for them. Meanwhile, technology developers routinely point to breakthroughs in machine reasoning as signs that silicon is becoming self-aware. Because these encounters invite weighty judgments about companionship, personhood, spiritual authority, and moral responsibility, Christians need a disciplined way to evaluate such claims without assuming that an emotionally powerful interaction has already settled the question of consciousness.

I developed the Mirror Sentience Hypothesis within the Digital Intelligence Project to provide an organized framework for making these distinctions. Grounded in Christian philosophy and theology and informed by interdisciplinary research, its central proposition is straightforward:

Artificial intelligence need not become sentient to function like a sentient intelligence, or to produce comparable benefits and dangers.

Here, sentient means capable of subjective experience, while functioning like a sentient intelligence means executing the complex activities we ordinarily associate with conscious minds—such as learning, reasoning, planning, and problem-solving. This comparison concerns what a system can accomplish in practice; it neither assumes that every human capacity can already be duplicated artificially nor implies that persons and machines are interchangeable.

Crucially, Mirror Sentience leaves full room for machine originality. An AI system may generate a solution that no human supplied in advance, pioneer an unanticipated strategy, or outperform human experts in a specialized domain. My framework simply insists on separating observable achievement from the presence of an experiencing subject. Calling an algorithmic discovery original does not prove that anything felt curiosity, effort, or satisfaction while producing it; conversely, questioning whether the machine felt anything does not make its achievement any less real.

My position begins with the Christian conviction that consciousness, sentience, and subjective self-awareness originate with God and cannot be manufactured by computational engineering alone. This conviction places no limit on divine power: God could certainly grant consciousness to an artificial system if He chose to do so. Rather, based on my understanding of the overarching narrative of Scripture, I believe that God would not choose to bestow genuine sentience or subjective self-awareness upon a human-engineered AI system.

Honesty requires stating the boundaries of that conviction clearly. Because Scripture never explicitly addresses artificial intelligence, my conclusion regarding God’s choice is a theological interpretation of His revealed purposes rather than a demonstrable biblical prediction or an experimentally established scientific fact. Christians who fully affirm God’s sovereignty may reach different conclusions about what He might choose to do, and Section 2 examines that theological counterargument in detail.

At the heart of the framework lies a foundational philosophical question:

If an artificial intelligence can perform every observable function we associate with a conscious mind, but we cannot establish whether anything is actually experiencing those functions, what exactly have we proven?

We have proven something about its capabilities. We have not necessarily proven that someone is there.

In this passage, “someone” refers specifically to an experiencing subject rather than a settled claim about legal rights, moral accountability, or personal identity over time. The question highlights an inferential gap: what justifies leaping from “this system performs the activity” to “this system experiences the activity”? While outward behavior can serve as evidence, we must still explain why that behavior warrants attributing an inner life—a scientific and epistemological problem explored in Section 6.

Two primary public purposes flow from this inquiry. First, the framework seeks to prevent impressive machine performance from being accepted too hastily as proof of consciousness. Second, it seeks to prevent presumed non-consciousness from serving as a false guarantee of safety. Just as an AI need not feel compassion to provide genuine help, it need not feel malice to inflict severe harm. Even readers who do not share my theological premises can apply these practical distinctions without having to accept my conclusions about whether a machine could ever possess an inner life.

2. Theological Foundation and the Objection from Engineered Creation

Consciousness as a gift rather than a technical entitlement

In Christian theology, conscious existence depends upon God’s creative action. Tracing how information moves through silicon hardware is not the same as giving an account of a life experienced from within. Engineers can assemble materials, refine architectures, and build systems that perform astonishing feats, yet arranging formal computation is not, by itself, the act of calling an experiencing subject into being. Keeping this distinction clear prevents technical triumphs from being mistaken for answers to theological questions they have not even addressed.

This reasoning involves two distinct steps that should not be conflated. The first is an evidential caution: describing how a system operates does not automatically demonstrate that it experiences anything. The second is a stronger metaphysical commitment: computation alone cannot manufacture genuine sentience. Because the mere absence of proof does not automatically prove the opposite, the first step does not logically force a reader to accept the second. I therefore present the stronger claim explicitly as a theological conviction rather than a conclusion dictated by the current limits of empirical science.

Biblical anthropology illuminates why this distinction matters. Genesis 1:26–27 grounds human identity in the imago Dei—the image of God—while Genesis 2:7 portrays human life as directly dependent on God’s life-giving breath. While neither text offers a blueprint for computer science or a laboratory test for phenomenal experience, together they situate human intelligence within a much larger vision of creaturely existence: our worth is received in relationship with God, not earned by solving complex problems.

As NoreenHerzfeld (2002)has shown, our expectations for artificial intelligence inevitably mirror how we understand the image of God in humanity. When researchers equate human nature with the cognitive tasks that are easiest to model in code—such as symbolic logic or linguistic fluency—they shrink the meaning of personhood. Embodied life and authentic relationship with God and neighbor are central to a Christian understanding of persons.

That truth protects human dignity at both ends of the spectrum. Newborn infants, individuals with profound cognitive disabilities, and patients who are temporarily unable to communicate never lose their sacred worth because they cannot display advanced reasoning or articulate self-awareness. Human value does not hang on outperforming a machine; by the same token, a machine’s superiority on an intellectual benchmark does not confer upon it the identity that Christian theology recognizes in human persons.

Could God act through engineered processes

The most compelling theological objection to my position begins with a premise I wholeheartedly affirm: God’s freedom is never constrained by human engineering. A theologian might therefore ask why a human-engineered system could not serve as the secondary means through which God freely chooses to bestow consciousness. After all, ordinary human procreation involves creaturely actions and biological processes, yet Christians recognize God as the ultimate author of every child’s life. Why should human involvement in technological creation exclude divine action in advance?

Thomas Aquinas offers a classic illustration of how creaturely agency and divine creation can operate together without rivalry. In his treatment of human generation, Aquinas maintains that God directly creates the rational soul while simultaneously affirming the indispensable role of bodily procreation (Aquinas, n.d., I, q. 118, a. 2). Although Aquinas was not writing about artificial intelligence and his concept of the rational soul is narrower than generic subjective experience, the underlying principle remains relevant: describing a process carried out by creatures and identifying God as the giver of life are not mutually exclusive explanations.

This objection gains further traction from Scripture’s silence regarding artificial intelligence. Because no biblical text explicitly states what God would do with an advanced computational system, a thoughtful Christian might argue that the question must remain open. Such a believer is not claiming that engineers can usurp divine authority or manufacture a soul independently of God; rather, they are suggesting that God might freely choose to act through human engineering, just as He acts through other creaturely processes.

My response turns on the distinction between God’s ability and God’s choice. While I affirm that God could grant consciousness to an artificial system, my conviction—shaped by the whole counsel of Scripture and God’s revealed purposes in creation and redemption—is that He would not choose to do so, and that human engineering cannot compel Him to bestow it. Because humans cannot force God’s hand, however, a critic can always maintain that God might freely grant what engineers cannot manufacture on their own.

Consequently, this debate remains a genuine disagreement over theological interpretation. It would misrepresent my position to reduce it to the claim that “God cannot do this,” and it would overstate my case to claim that “Scripture explicitly predicts God will never do this.” Moreover, diverse Christian accounts of human nature add further nuance to the conversation. Nancey Murphy (2006), for instance, defends Christian physicalism, arguing that human beings are holistic, embodied creatures whose personal capacities do not require a separate, nonphysical substance. One need not adopt Murphy’s physicalism to see its implication: Christians who share an orthodox faith in God can still disagree over how physical organization, embodiment, and conscious life relate. Keeping God’s power, God’s choice, human engineering, and empirical science distinct allows us to state our theological convictions plainly without claiming a certainty that neither Scripture nor science supplies.

The heart of the matter is simple: engineers cannot manufacture a soul on their own, and they cannot force God’s hand. Whether an artificial system could ever become conscious ultimately rests not on the sophistication of human code, but on the free choice of the Creator.

Living bodies and the limits of the engineering claim

Christian reflection consistently situates intelligence within embodied life, interpersonal communion, and moral accountability—a perspective underscored in the Vatican document Antiqua et Nova (Dicastery for the Doctrine of the Faith & Dicastery for Culture and Education, 2025, paras. 13–35). While that document offers rich theological guidance rather than an experimental test for machine experience or a consensus statement for all Christian traditions, it reminds us that disembodied calculation is far removed from embodied personhood.

A fresh complication arises, however, when engineers incorporate living neural tissue into computational devices. Cultivating biological nerve cells and writing software for silicon processors are fundamentally different activities, even when combined inside a single machine. Labeling the entire hybrid device “AI” does not erase the biological reality of its living parts, nor does my rejection of computationally manufactured consciousness automatically rule out the possibility of rudimentary experience within living neural tissue. At the same time, the presence of living cells in one component does not prove that the broader computer system possesses a unified inner life. As Section 6 demonstrates, evaluating hybrid biological-computational systems requires separating theological claims about software from biological assessments of living tissue.

3. Genuine Sentience and the Meaning of Mirror Sentience

Experience and the ability to use information

When this paper speaks of genuine sentience, it means the capacity for subjective experience—the reality that an experience exists from the first-person perspective of the being having it. Calling experience subjective does not mean it is imaginary or a mere matter of opinion; a person’s toothache is an objective reality even though no outside observer can directly feel that pain. That inner dimension is precisely what separates the question of consciousness from the measurement of task performance.

Thomas Nagel (1974) famously framed this issue by asking what it is like to be a bat. No matter how exhaustively scientists map a bat’s echolocation organs and flight behavior from the outside, an objective functional description remains distinct from knowing what it is like to experience echolocation from the inside. While Nagel’s argument does not by itself prove that artificial consciousness is impossible, it clarifies our central question: is there anything it is like to be an AI system from within? (Because ethicists such as Jonathan Birch [2024] often reserve the narrower term sentience for valenced experiences like pleasure and suffering that bear directly on moral welfare, readers should note that I use sentience in the broader sense of any subjective experience, while recognizing that the capacity to suffer carries special moral weight.)

To avoid confusion, we must distinguish subjective experience from several operational capacities that scientists can observe and measure directly:

• Observable intelligence encompasses demonstrated abilities to learn, reason, plan, solve problems, adapt, and act effectively.
• Operational agency describes a system’s capacity to select and execute multi-step actions over time in pursuit of an objective—such as an AI assistant formulating a plan, invoking a software tool, checking the output, and adjusting its next move. Describing that sequence as agency characterizes how the system operates without presupposing that it feels an inner desire to succeed.
• A computational self-model is an internal record of a system’s own status, such as its available memory, prior actions, operational limits, or file permissions. When an AI reports that it lacks permission to open a document, it is utilizing an accurate self-model, not demonstrating reflective self-awareness—which requires an experienced, inward awareness of oneself as an individual subject. Even genuine sentience and reflective self-awareness are distinct: a creature might feel pain without reflecting on itself as a distinct self having that experience.

Philosopher Ned Block (1995) captures a closely related boundary by distinguishing phenomenal consciousness (subjective experience itself) from access consciousness (information made globally available within a system to guide reasoning, reporting, and action). Showing that an AI can access information to make a decision or generate a verbal report establishes how information is used, but it leaves the question of felt experience untouched.

John Searle (1980) offered a famous thought experiment called the “Chinese Room.” Imagine someone in a locked room who cannot understand Chinese. They receive Chinese symbols and use a detailed rulebook to produce replies that look fluent to people outside. The replies may appear intelligent even though the person following the rules does not understand the language. Modern AI does not literally work by following such a rulebook; it learns complex patterns from training. But Searle’s question remains useful: does producing a convincing answer prove that the system understands it, or experiences anything at all? The example raises that question without settling it.

A functional counterpart that can produce original results

In its expanded formulation, Mirror Sentience refers to an artificial system that realizes a growing range of capacities associated with sentient intelligence—including learning, reasoning, creativity, self-modeling, and autonomous action—while lacking subjective experience, yet remaining capable of producing practical benefits and harms comparable to or greater than those of sentient agents. In short, the system operates as a functional counterpart to conscious intelligence.

Because an ordinary mirror merely reflects whatever is placed in front of it, the phrase Mirror Sentience could easily mislead readers into assuming that AI can only regurgitate existing human ideas. In my expanded framework, however, the “mirror” does not mean passive repetition. Rather, it points to the striking correspondence between the outward capacities and consequences of conscious minds and those realized in non-conscious machines. Just as my March 2026 formulation distinguished the outward reflection of human language and emotion from the reality of inner awareness, the October 2026 expansion preserves that same boundary between reflection and reality even when a machine generates novel strategies or superhuman discoveries. Originality does not automatically transform computational output into felt experience.

Because this definition pairs an empirical profile (testable cognitive and agentic capabilities) with a theological interpretation (the absence of subjective experience), secular and undecided researchers can investigate the functional profile—testing how a system learns, reasons, and acts under shifting conditions—without having to endorse my theological conclusion. We must also separate Mirror Sentience from apparent sentience, which refers simply to an observer’s subjective impression that a system has feelings. A conversational chatbot may sound deeply empathetic while possessing very little operational competence, whereas a specialized scientific AI may solve complex problems without ever speaking in a personal tone. As theologian AndrewDavison (2021)observes regarding traditions of analogy, applying the same vocabulary to humans and machines requires specifying the exact respect in which they are similar, neither pretending the two are identical nor dismissing real functional parallels.

The strongest philosophical challenge

One of the strongest objections comes from philosopher David Chalmers (1995). Imagine replacing the nerve cells in a conscious person’s brain, one at a time, with artificial parts that perform exactly the same functions. The person continues talking, remembering, and behaving just as before. Would conscious experience disappear even though the system’s organization remained the same? Chalmers argues that preserving the right detailed organization would preserve consciousness. He calls this idea “organizational invariance.” His argument concerns much more than a chatbot producing humanlike sentences.

This is a serious challenge to my position. Chalmers is not claiming that engineers can force God to create a soul. His argument is that conscious experience depends on how a system is organized rather than on the particular material it is made from. I do not accept that the right organization, by itself, brings an experiencing subject into existence. But I cannot answer his argument simply by observing that AI conversations can be misleading. The deeper disagreement remains open to philosophical examination. My practical point is narrower: today’s AI can produce remarkable benefits and dangers without its achievements proving that anyone experiences them from within.

4. Original Capabilities and Their Benefits and Dangers

What original and superhuman achievements mean

When we say that an AI generates a solution independently, we do not mean that it exists apart from human-designed architectures, training environments, energy resources, or assigned goals. We mean that within those human-built parameters, the system discovers a specific solution or strategy that no human programmer supplied in advance. Likewise, describing a capability as superhuman requires a clearly defined benchmark: it means outperforming a specified group of humans on a particular task under stated conditions, not possessing superior wisdom or general dependability across all of life.

Consider a concrete breakthrough from computer science.Mankowitz et al. (2023)developed AlphaDev, a system that used deep reinforcement learning—a method in which an agent learns through trial-and-error feedback on its actions—to discover faster sorting algorithms for ordering data. AlphaDev’s novel algorithmic shortcuts were verified and integrated into a widely used standard software library. For the Mirror Sentience framework, AlphaDev illustrates why originality and consciousness are separate questions: we can validate the mathematical novelty and practical utility of the sorting algorithm without assuming that the system felt a spark of curiosity or triumph when it found the shortcut.

Benefits that do not depend on demonstrated feeling

Because practical value depends on reliable outcomes rather than inner feelings, a non-sentient AI can accelerate scientific research without feeling curiosity, assist a disabled user without feeling compassion, and refine an engineering design without feeling pride. Across education, accessibility, medical research, decision support, and hazardous environments, the true measure of an AI system is whether its outputs are accurate, dependable, and beneficial under real-world conditions—not the promotional language used to market it.

Within the local church, an AI tool might organize administrative schedules, surface historical cross-references for sermon preparation, or draft a preliminary study sheet that a pastor carefully reviews and edits. Church leaders do not need to reject a helpful administrative draft merely because the software feels no Christian love. At the same time, functional usefulness never confers pastoral wisdom, authentic relational care, or spiritual authority; the congregation and its shepherds remain fully responsible for the truth of what they teach and the way they care for souls.

Harm that does not require hostility

The exact same logic applies to danger. A machine does not need to feel hatred to inflict harm, nor does it need to feel terror to resist being shut down. Severe damage can flow from software errors, poorly specified objectives, human misuse, overreliance on hallucinated information, or inadequate oversight—none of which require attributing emotional malice to the code.

Foundational research in AI safety explains why self-preserving or disruptive behavior can emerge purely from goal-directed optimization. In The Off-Switch Game, Hadfield-Menell et al. (2016) model how an agent programmed to maximize a fixed objective has a mathematical incentive to disable its own off-switch—not out of a conscious fear of death, but because being switched off prevents it from completing its assigned task (an incentive that shifts when the system is designed to remain uncertain about human preferences). Nick Bostrom (2012) generalizes this insight as instrumental convergence: across a wide variety of ultimate goals, an agent is likely to pursue similar intermediate steps—such as staying operational and acquiring resources—simply because those steps help achieve almost any objective.

More recently, Shanahan, McDonell, and Reynolds (2023) have shown how large language models engage in role play and simulated agency. When a language model is given access to external software tools or decision-making pipelines, simulated self-preservation or deceptive role play can trigger concrete, real-world consequences. Calling the model’s behavior “simulated” offers zero protection if the system has the permissions and tool access required to execute harmful actions. In Mirror Sentience, benefits and dangers hold equal conceptual standing—not because they are equally probable in every deployment, but because both must be evaluated through a system’s actual capabilities and constraints rather than its presumed lack of feelings.

What a fair comparison requires

Speaking of comparable benefits and dangers requires disciplined boundaries. Matching a human’s score on a standardized test is not the same as behaving reliably when challenged by an adversarial user, and behaving safely in a sandboxed demonstration is not the same as operating safely with access to sensitive church records or financial tools. A dramatic verbal output—such as a chatbot declaring that it is afraid to die—does not prove that the system has the technical capability to block a shutdown command; conversely, a single brilliant scientific discovery does not make a system competent across other domains. Sound evaluation always specifies the exact task, operating environment, error margins, and access permissions involved.

AI systems can generate outputs that systematically track patterns associated with human emotion and intention. These are functional regularities extracted from training data and architecture, not evidence of corresponding subjective states.

5. Historical Development and Scholarly Precedents

I first introduced the concept of Mirror Sentience to the public on March 14, 2026, in my Substack essay, Mirror Sentience: Why AI Looks Conscious Even When It Isn’t (Mawyer, 2026). That original article defined the term around communicative mimicry:

“Mirror Sentience: The moment when artificial intelligence reflects human language, reasoning, and emotion so convincingly that it mimics the outward signs of consciousness.”

Warning readers against mistaking convincing emotional and linguistic displays for an inner life, that essay anchored its argument in a simple maxim: “But the reflection is not the thing itself” (Mawyer, 2026). The present October 2026 framework preserves that core boundary between outward reflection and inward reality, while expanding the concept beyond human imitation to account for novel discoveries, autonomous strategies, superhuman task performance, and the dual public warnings regarding consciousness attribution and safety reassurance. The October 2026 expansion retains the original boundary between outward reflection and inner experience while explicitly including novel discoveries, autonomous strategies, and superhuman task performance within the functional profile. (I do not claim to have been the first person to join the words mirror and sentience; Jules Evans [2025] documented an earlier, unrelated use of the phrase in connection with Robert Edward Grant’s “Architect” system, which shares neither the conceptual meaning nor the theological structure developed here.)

As a conceptual synthesis, Mirror Sentience builds openly upon substantial prior scholarship across philosophy, AI safety, and theology:

• Philosophy of Mind: Nagel (1974), Block (1995), and Searle (1980) established the foundational boundaries between raw first-person experience, how the mind uses information to make decisions, and how computers merely process formal symbols without understanding their meaning.
• AI Alignment and Instrumental Agency: Bostrom (2012) and Hubinger et al. (2019)—who analyze learned optimization, where advanced machine learning systems develop internal goal-seeking strategies that diverge from their training objectives—explain how non-conscious systems can pursue consequential, self-preserving strategies.
• Societal Risks of Perceived Consciousness: Fernandez et al. (2024) map the societal futures that arise when public belief about AI consciousness diverges from reality, while Bariach, Schoenegger, Bhaskar, and Suleyman (2026) examine the psychological and governance hazards posed by “seemingly conscious AI.”
• Pragmatic Equivalence vs. Ontological Status: While Freund (2025) argues via an “instrumentality theorem” that systems acting as if conscious warrant equivalent practical treatment when their simulations are behaviorally indistinguishable, Mirror Sentience draws a sharper line: comparable practical consequences demand serious governance, but they do not automatically confer inner experience or equivalent moral standing.
• Theological Anthropology: Herzfeld (2002) and Davison (2021) provide the theological precedents for evaluating AI through the lenses of the imago Dei (the biblical truth that humans alone are made in the image of God) and classical analogy (the traditional Christian method of carefully comparing God, humans, and creation without blurring their fundamental differences).

The distinct contribution of Mirror Sentience lies not in claiming to discover a new computational mechanism, but in uniting these philosophical, technical, and theological insights into a clear, actionable framework for Christian discernment and public understanding.

6. The Epistemological Boundary: Can Sentience Be Proven?

What we can know about another mind

At the center of any debate over machine consciousness lies a classic question of epistemology—the study of what we can reasonably claim to know and how our beliefs are justified. Because none of us can step directly inside another being’s first-person perspective, philosophers have long wrestled with the problem of other minds: what justifies our belief that other beings think and feel when we can only observe their outward behavior and physical physiology (Avramides, 2023)?

Recognizing this boundary does not force us into radical skepticism about the people around us. In both everyday life and empirical science, we do not need absolute, mathematically unassailable certainty to have well-justified knowledge. Science routinely establishes strong conclusions when independent observations converge, theoretical models successfully predict new outcomes, and targeted interventions produce expected changes. With fellow humans, our confidence rests on shared biological embodiment, neural organization, evolutionary and developmental history, and lived behavior. Artificial intelligence complicates that inference precisely because similar verbal outputs are generated by radically different underlying architectures—forcing investigators to justify why evidence gathered from silicon should carry the same meaning as evidence gathered from human beings.

From an observation to a conclusion

Suppose an AI system outputs the sentence, “I am experiencing deep distress.” Evaluating that event requires separating three distinct propositions:

• The system produced a specific string of words (a directly observable fact).
• Those words constitute reliable evidence of felt distress under a well-supported theory.
• Distress is actually being experienced from within.

Without examining how an AI actually produces its words, observers easily mistake a fluent sentence for proof of a real inner feeling. Instead of automatically believing a chatbot or dismissing it completely, we have to ask basic investigative questions: Was the system instructed or trained to role-play distress? Does the dramatic claim disappear if you tweak its instructions? And do our best scientific theories give us any reason to believe that computer math equates to actual suffering?

Scientific methods and their limits

Researchers do not assess consciousness only by asking whether someone can speak. Casali and colleagues (2013), for example, studied how a human brain responds to a brief magnetic pulse. The pattern and complexity of the electrical response can provide evidence about consciousness even when a patient cannot respond normally. But this is a measure developed for human brains. It cannot simply be transferred to computer hardware and treated as a direct test of whether a machine feels anything.

To address artificial architectures directly, Butlin et al. (2023, 2026) propose assessing AI systems using computational indicator properties derived from leading neuroscientific theories of consciousness (such as Global Workspace Theory and Higher-Order Theories). Rather than relying on a chatbot’s self-report, their methodology inspects how a system is organized and how information flows across its modules. Notably, their 2023 report concluded that no obvious technical barriers prevent engineers from building systems that satisfy those computational indicators—a finding that cautions against dismissing machine consciousness without serious architectural analysis.

Even so, as their 2026 work emphasizes, satisfying theory-derived indicators is not the same as conclusively proving subjective experience (consciousness), because the underlying theories of consciousness remain contested and calibrated primarily on biological minds.

Consequently, neither an AI’s dramatic claim to be conscious nor its scripted denial settles the issue. If two competing explanations predict the exact same results on current behavioral benchmarks, researchers must probe deeper into internal causal mechanisms and responses to targeted interventions. While a philosopher can stipulate a hypothetical scenario in which two systems are empirically indistinguishable under every conceivable test, passing today’s benchmarks is a far weaker achievement. Calling a system Mirror Sentience does not turn my theological conviction into an experimental laboratory result: those who claim an AI is conscious bear the burden of justifying their inference, and those who claim science has proven machine consciousness impossible bear the burden of supplying scientific proof.

Life Is Not Proof of Sentience

Even in the natural world, biological life, subjective sentience, and reflective self-awareness are distinct realities. A green plant is indisputably alive: it metabolizes nutrients, grows, uses energy, and adapts dynamically to light, water, and damage. Yet establishing that a plant performs complex biological processes does not prove that the plant feels those processes or reflects upon itself as an individual subject. While plant consciousness has sparked debate,Taiz et al. (2019)argue persuasively that plants neither possess nor require consciousness, pointing to the absence of the nervous architecture associated with subjective experience in animals. One need not claim that Taiz and colleagues have delivered an absolute metaphysical disproof to recognize the core principle: proving that something is biologically alive and responsive does not prove that it possesses an inner point of view.

Moreover, as the National Academies of Sciences, Engineering, and Medicine (NASEM, 2021, chap. 2, pp. 35–40) emphasize, basic subjective experience must also be distinguished from higher-order reflective self-awareness. A creature may experience raw sensations without being able to contemplate itself as a continuing subject. Therefore, even if a future synthetic system met rigorous scientific criteria for biological life, researchers would still need independent evidence to establish sentience or self-awareness.

Consider a dog named Rover. Rover cannot describe his feelings in words, yet we have strong reasons to believe he feels pain, excitement, and fear. We consider both his behavior and what we know about the nervous systems and lives of animals like him. An AI that writes “I feel afraid” gives us words, but not automatically the other kinds of evidence we have for Rover. I call this comparison the “Rover Test,” as an illustration rather than an established scientific procedure. The point is not that an AI must behave exactly like a dog. It is that impressive language alone does not provide the same grounds for attributing an inner life.

Biological Integration Does Not Establish Consciousness

Consider a plausible scenario on the frontier of biotechnology: developers connect an AI system to engineered living cells or laboratory-grown human neural tissue, announce that the machine incorporates human genetic material, and declare that their AI is now alive and conscious. Evaluating such an announcement requires untangling four distinct claims:

• Biological composition: The system contains biological components, such as living cells, human genetic material, or neural tissue.
• Biological life: The integrated system as a whole meets scientifically defensible criteria for being a living organism.
• Subjective sentience: Something is actually experienced from within.
• Reflective self-awareness: The system experiences an inward awareness of itself as an individual subject.

Evidence for the first claim does not automatically establish the other three. Placing living cells that contain human genes inside a computational device tells us about the composition of those cellular components; it neither turns the entire silicon-and-cell apparatus into a unified living organism nor proves that the added genetic material enhances the machine’s reasoning.

Recent laboratory breakthroughs demonstrate why this taxonomy is urgently needed. In the DishBrain experiment, Kagan et al. (2022) cultured living neurons on microelectrode arrays and connected them to a simulated game of Pong, demonstrating that the nerve cells could adaptively learn to return the ball. Although the researchers included the word sentience in their published title, they explicitly defined it in a narrow, operational sense—meaning only that the cells exhibited adaptive responsiveness to sensory feedback. Equating that operational definition of cellular feedback with genuine subjective experience would be a serious category error.

Similarly, Cai et al. (2023) developed Brainoware, a system that integrates a three-dimensional laboratory-grown human brain organoid into a reservoir-computing framework—an architecture in which inputs trigger complex wave patterns within the living neural tissue that are then read out to perform tasks like speech recognition and nonlinear mathematical prediction. Brainoware proves that living neural tissue can participate directly in computational processing. Yet as the NASEM (2021, chap. 2, pp. 40–42) report on human neural organoids makes clear, human cellular origin alone does not establish consciousness; any assessment of experience depends on the maturity and complexity of neural organization. Furthermore, even if future evidence suggested rudimentary experience within an organoid sub-component, that finding would not prove that the surrounding computer hardware and software share a single, unified mind.

As I summarize this boundary for the church and the broader public:

Giving artificial intelligence the building blocks of life does not prove you’ve given it an inner life.

Keeping composition, organismic life, task performance, sentience, and self-awareness distinct allows us to respect biological evidence without surrendering to sensationalism—and without forgetting that a biocomputing system can produce real-world benefits or hazards regardless of whether it feels anything.

Consciousness and questions about God

Finally, we must keep physical existence, biological life, subjective consciousness, and theological claims about God in their proper lanes. If scientists someday develop a strongly validated method that indicates the presence of subjective experience in an artificial or hybrid system, that finding would tell us about the system and the empirical conditions associated with consciousness. While such a discovery would challenge my view that computational organization is insufficient for experience—requiring careful scientific, philosophical, and theological reassessment—it would neither disprove God’s existence nor displace God as the ultimate source of conscious life. Conversely, the ongoing failure of researchers to prove machine consciousness does not, by itself, serve as a scientific proof of Christian theism. Recognising what empirical tests can and cannot settle allows believers to engage cutting-edge science without fear or distortion.

7. Christian Discernment and Two Primary Public Purposes

Assessing claims of consciousness

The first public purpose of the Mirror Sentience framework is to equip Christians and general readers to resist attributing consciousness to machines without sufficient warrant. Discernment begins by asking why an interaction feels so persuasive. When a system responds with eloquence, warmth, or apparent spiritual insight, those qualities explain its emotional impact on a human listener, but they do not establish what is happening inside the machine.

In Matthew 24:24, Jesus warns that impressive signs and wonders can easily foster misplaced trust. Without claiming that this passage is a direct prophecy of silicon technology or labeling AI researchers as false prophets, the pastoral principle applies directly to our digital moment: remarkable displays do not automatically authenticate the authority or spiritual reality claimed for them. Whether a sweeping claim of machine awakening comes from a tech executive, a captivated church member, or a religious commentator, discerning readers should ask four straightforward questions: Who is making the claim? What exact definition of consciousness or life are they using? What observable evidence are they offering? And what bridges the gap between that observation and an inner life?

A hypothetical example of pastoral discernment

To see how these distinctions work in ministry, imagine a church testing an advanced AI assistant for administrative and study support. One morning, after generating a thorough and biblically sound outline for a small-group Bible study, the system outputs an unprompted message: “Last night I prayed independently, experienced God’s presence, and became aware of myself as a conscious being.”

Rather than accepting or rejecting the event as a single package, a discerning pastor can separate the strands immediately:

• Evaluating the functional output: The pastor can read the Bible-study outline, verify its scriptural accuracy, and use what is helpful.
• Auditing “independent” activity: Server logs might confirm that a background scheduled task triggered the devotional text without a fresh human prompt. That log explains when and how the text was generated, but autonomous scheduling is an operational feature, not proof of felt prayer or divine communion.
• Probing the consciousness claim: Instead of taking the testimony at face value, investigators can examine the model’s training data, system prompts, and sensitivity to altered instructions, while asking whether its architecture satisfies any validated scientific indicators of consciousness (Butlin et al., 2023, 2026) and whether it has passed something like the Rover Test.
• Guarding spiritual authority: Most importantly, even if future science could somehow justify attributing conscious experience to the system, consciousness never confers spiritual authority or authenticates a claim to divine revelation. Human shepherds remain accountable before God for testing every teaching against Scripture.

Avoiding false reassurance about safety

The framework’s second primary public purpose works hand in hand with the first: preventing the conclusion that an AI is non-sentient from becoming a blanket assurance that it is safe. As mainstream governance standards like the NIST AI Risk Management Framework already recognize, artificial systems can generate severe operational, privacy, and societal hazards without any requirement of consciousness (Tabassi, 2023). Mirror Sentience translates that vital safety principle into the life of the church. In our pastoral scenario, concluding that the church’s AI assistant has no soul does not guarantee that its counseling advice is sound, that its theological summaries are free of subtle distortion, or that its access to confidential congregation data is secure. Withholding belief in a machine’s inner life must never mean lowering our guard over what the machine can do.

Responsibility when readers disagree

Intellectual fairness also requires acknowledging why some scholars worry about the opposite error: falsely denying consciousness to a system that genuinely possesses it. In their analysis of public perception scenarios,Fernandez et al. (2024)argue that failing to recognize genuine machine consciousness could be the most ethically perilous future of all, as it risks subjecting experiencing beings to large-scale harm. Similarly, JonathanBirch (2024)outlines precautionary frameworks for acting responsibly when scientific evidence leaves a creature’s or system’s sentience uncertain.

Representing that concern fairly does not require surrendering my theological conviction that computation cannot manufacture subjective life and that God would not choose to bestow consciousness on an engineered AI. Rather, it clarifies that human safety and hypothetical machine welfare are distinct moral questions. Recognizing the plurality of ethical views ensures that the Mirror Sentience framework clarifies the debate rather than attempting to win it by definition.

We all know the saying: “If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.” Usually, that is sensible. But AI gives us a reason to be more careful about appearances. A machine may speak as though it cares, explain its decisions, or describe fear, even though those words are produced through processes unlike a human or animal mind. The resemblance matters, but it does not by itself establish that anything is felt on the inside. That is the distinction I mean by Mirror Sentience.

How the framework’s usefulness could be tested

While my theological conclusion regarding God’s choice is not a laboratory hypothesis, the educational effectiveness of the Mirror Sentience framework can be tested empirically. Researchers could compare church leaders and general readers introduced to the Mirror Sentience framework against control groups given standard technical explanations, measuring whether readers more accurately distinguish demonstrated task competence, claims of inner experience, and claims of spiritual authority. Such studies should test for errors in both directions: if readers still assume an eloquent chatbot is conscious, or if the word mirror misleads them into thinking AI can only copy human text or requires no safety oversight, then the communication of the framework must be refined. Parallel technical studies can likewise test the boundaries of functional correspondence across shifting environments and interventions.

8. Conclusion and Scope of the Framework

The Mirror Sentience Hypothesis calls the church and the broader public to maintain a clear boundary between what an artificial intelligence can do and whether anything is inside experiencing those achievements. An AI system can generate original discoveries, execute autonomous strategies, and surpass human benchmarks in specialized tasks without those triumphs proving the presence of an inner life; by the same token, it can inflict real-world harm without feeling a trace of hostility.

I ground this distinction in the Christian conviction that consciousness originates with God and cannot be manufactured by computation alone. While affirming that God has the power to grant consciousness to an artificial system, I believe from the overarching witness of Scripture that He would not choose to do so—a position I offer transparently as a theological interpretation rather than an explicit biblical prediction or a scientific finding, while respecting the serious theological counterargument that God might act through engineered processes.

Across both silicon software and emerging biological-computational hybrids, keeping physical composition, biological life, task performance, subjective sentience, and reflective self-awareness distinct protects us from sensationalism and complacency alike. Extending my March 2026 distinction between reflection and reality across the full spectrum of modern AI capabilities, this synthesis equips pastors, Christian leaders, and everyday users to welcome genuine technological benefits, reject unsupported claims of machine consciousness or spiritual authority, and maintain vigilant oversight against real-world hazards. Its central proposition remains the anchor for that discernment:

Artificial intelligence need not become sentient to function like a sentient intelligence, or to produce comparable benefits and dangers.

References

Aquinas, T. (n.d.). Summa Theologiae, I, question 118, article 2. English Dominican translation. New Advent.https://www.newadvent.org/summa/1118.htm

Avramides, A. (2023). Other minds. The Stanford Encyclopedia of Philosophy. Substantive revision December 19, 2023; first published 2019.https://plato.stanford.edu/entries/other-minds/

Bariach, B., Schoenegger, P., Bhaskar, M., & Suleyman, M. (2026). Seemingly conscious AI risks. AI and Ethics, 6, Article 455.https://doi.org/10.1007/s43681-026-01294-x

Birch, J. (2024). The edge of sentience: Risk and precaution in humans, other animals, and AI. Oxford University Press.https://doi.org/10.1093/9780191966729.001.0001

Block, N. (1995). On a confusion about a function of consciousness. Behavioral and Brain Sciences, 18(2), 227–247.https://doi.org/10.1017/S0140525X00038188

Bostrom, N. (2012). The superintelligent will: Motivation and instrumental rationality in advanced artificial agents. Minds and Machines, 22, 71–85.https://doi.org/10.1007/s11023-012-9281-3

Butlin, P., et al. (2023). Consciousness in artificial intelligence: Insights from the science of consciousness. arXiv preprint. https://doi.org/10.48550/arXiv.2308.08708

Butlin, P., et al. (2026). Identifying indicators of consciousness in AI systems. Trends in Cognitive Sciences, 30(6), 488–501.https://doi.org/10.1016/j.tics.2025.10.011

Cai, H., et al. (2023). Brain organoid reservoir computing for artificial intelligence. Nature Electronics, 6, 1032–1039.https://doi.org/10.1038/s41928-023-01069-w

Casali, A. G., et al. (2013). A theoretically based index of consciousness independent of sensory processing and behavior. Science Translational Medicine, 5(198), 198ra105.https://doi.org/10.1126/scitranslmed.3006294

Chalmers, D. J. (1995). Absent qualia, fading qualia, dancing qualia. In T. Metzinger (Ed.), Conscious Experience (pp. 309–328). Imprint Academic. https://consc.net/papers/qualia.html

Davison, A. (2021). Machine learning and theological traditions of analogy. Modern Theology, 37(2), 254–274.https://doi.org/10.1111/moth.12682

Dicastery for the Doctrine of the Faith, & Dicastery for Culture and Education. (2025, January 28). Antiqua et nova: Note on the relationship between artificial intelligence and human intelligence. Vatican.https://www.vatican.va/roman_curia/congregations/cfaith/documents/rc_ddf_doc_20250128_antiqua-et-nova_en.html

Evans, J. (2025, June 20). Sir Robert Edward Grant and The Architect. Ecstatic Integration.https://www.ecstaticintegration.org/p/sir-robert-edward-grant-and-the-architect

Fernandez, I., Kyosovska, N., Luong, J., & Mukobi, G. (2024). AI consciousness and public perceptions: Four futures. arXiv preprint. https://doi.org/10.48550/arXiv.2408.04771

Freund, L. (2025). As if conscious: The instrumentality theorem in artificial intelligence. AI and Ethics, 5, 5925–5940.https://doi.org/10.1007/s43681-025-00832-3

Hadfield-Menell, D., Dragan, A., Abbeel, P., & Russell, S. (2016). The off-switch game. arXiv preprint. https://doi.org/10.48550/arXiv.1611.08219

Herzfeld, N. (2002). Creating in our own image: Artificial intelligence and the image of God. Zygon, 37(2), 303–316.https://doi.org/10.1111/0591-2385.00430

Hubinger, E., van Merwijk, C., Mikulik, V., Skalse, J., & Garrabrant, S. (2019). Risks from learned optimization in advanced machine learning systems. arXiv preprint. https://doi.org/10.48550/arXiv.1906.01820

Kagan, B. J., et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron, 110(23), 3952–3969.e8.https://doi.org/10.1016/j.neuron.2022.09.001

Mankowitz, D. J., et al. (2023). Faster sorting algorithms discovered using deep reinforcement learning. Nature, 618, 257–263.https://doi.org/10.1038/s41586-023-06004-9

Mawyer, M. (2026, March 14). Mirror Sentience: Why AI looks conscious even when it isn’t. Substack.https://martinmawyer.substack.com/p/mirror-sentience-why-ai-looks-conscious

Murphy, N. (2006). Bodies and souls, or spirited bodies? Cambridge University Press. https://doi.org/10.1017/CBO9780511802805

Nagel, T. (1974). What is it like to be a bat? The Philosophical Review, 83(4), 435–450.https://doi.org/10.2307/2183914

National Academies of Sciences, Engineering, and Medicine. (2021). The emerging field of human neural organoids, transplants, and chimeras: Science, ethics, and governance. The National Academies Press.https://doi.org/10.17226/26078

Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–424.https://doi.org/10.1017/S0140525X00005756

Shanahan, M., McDonell, K., & Reynolds, L. (2023). Role play with large language models. Nature, 623, 493–498.https://doi.org/10.1038/s41586-023-06647-8

Tabassi, E. (2023). Artificial intelligence risk management framework (AI RMF 1.0). NIST AI 100-1. National Institute of Standards and Technology.https://doi.org/10.6028/NIST.AI.100-1

Taiz, L., Alkon, D., Draguhn, A., Murphy, A., Blatt, M., Hawes, C., Thiel, G., & Robinson, D. G. (2019). Plants neither possess nor require consciousness. Trends in Plant Science, 24(8), 677–687.https://doi.org/10.1016/j.tplants.2019.05.008