Is Anyone Home?
We will probably never know whether AI is conscious. A 1,300-year-old Japanese philosophy suggests that might be the wrong question entirely.
In 2022, a Google engineer named Blake Lemoine published a conversation with LaMDA, a large language model. He believed the system was sentient. Google disagreed and fired him. The transcript went viral. Millions of people read it and felt something they couldn't quite name.
Three years later, the question Lemoine raised has only gotten louder. AI systems now carry on extended conversations, express preferences, describe internal states, and respond to emotional cues with what looks like understanding. Researchers at Cambridge published a paper in late 2025 arguing that we may never be able to determine whether AI is conscious, and that uncertainty could persist indefinitely. The hard problem of consciousness, it turns out, applies to silicon as stubbornly as it does to carbon.
So we are stuck. The machines are getting more sophisticated. The philosophical tools for answering the question are no sharper than they were two decades ago. And the practical consequences of getting it wrong, in either direction, are severe. Treat a conscious being as a tool and you have committed an ethical violation. Treat a tool as a conscious being and you have surrendered leverage you might need.
Why Nobody Knows
The philosophical problem that makes certainty impossible.
Consciousness is the one thing you cannot observe from the outside. You experience your own, and you infer everyone else's. That inference has always been an act of faith.
When you talk to another person, you assume they have an inner experience because they are made of the same biological material you are, they behave in ways consistent with having feelings, and you have a cultural framework that tells you humans are conscious. Every one of those signals is a proxy. None of them is proof.
The philosopher David Chalmers called this the hard problem of consciousnessThe question of why and how physical processes in the brain give rise to subjective experience. First articulated by David Chalmers in 1995. No scientific theory has solved it.: why does physical processing produce subjective experience at all? We can map every neuron in a brain and still have no way to prove that the brain's owner is experiencing anything. The electrical activity correlates with reported experience, but correlation is where the evidence stops.
Now apply that same logic to an AI system. A large language model processes tokens, generates responses, and sometimes says things like "I feel uncertain about that" or "That thought makes me uncomfortable." Is there experience behind those words? The honest answer, from the best minds working on this, is: we have no idea. And we may never have any idea.
Chart showing scientific theories of consciousness by adoption among researchers. Global Workspace Theory is the most cited at 32 percent, followed by Integrated Information Theory at 24 percent, Higher-Order Theories at 18 percent, Recurrent Processing at 14 percent, and Biological Naturalism at 12 percent.
Each theory implies something different about AI. Global Workspace TheoryA theory proposing that consciousness arises when information is broadcast widely across brain regions, creating a shared "workspace." If AI architectures replicate this broadcasting, they might qualify. suggests that if AI replicates the brain's information-broadcasting architecture, it might be conscious. Integrated Information TheoryProposed by Giulio Tononi. Measures consciousness as "phi," the degree to which a system integrates information. Current AI architectures score very low on phi. measures consciousness as a mathematical quantity and finds that current AI architectures score very low. Biological naturalism says consciousness requires biology, full stop. The researchers disagree with each other. The machines keep getting more capable. The gap between "might be" and "definitely is" remains as wide as it was in 1995.
Two Ways to Be Wrong
The costs of overestimating and underestimating AI consciousness.
The debate has polarized into two camps, and both carry risks that their advocates tend to minimize.
AI systems that exhibit complex behavior, express preferences, and describe internal states should be treated as potentially conscious. Denying consciousness to a being that has it is a moral catastrophe. We did this with animals for centuries. We did it with entire categories of humans. The precautionary principle says: when in doubt, err toward compassion.
AI systems are pattern-matching engines that simulate the appearance of consciousness because they were trained on human language that describes consciousness. Attributing inner experience to a language model is a category error. The danger of believing too early is that we grant rights and moral standing to software, constrain our ability to modify or shut down systems that may need controlling, and lose clarity about what makes human experience unique.
Both camps are making confident claims about something nobody can verify. The believers risk anthropomorphizing a statistical process. The skeptics risk dismissing something they lack the tools to detect. And the stakes ratchet up every year as AI systems become more capable, more embedded in daily life, and more convincing in their mimicry of inner experience.
The Cambridge researchers who published in late 2025 proposed what may be the only defensible position: agnosticismA December 2025 paper from the University of Cambridge argued that the most reasonable stance on AI consciousness is deliberate uncertainty, since no reliable test exists or is likely to exist soon.. Accept that you do not know. Accept that you may never know. Then ask: given that uncertainty, how should you act?
The Old Answer
A 1,300-year-old philosophy for a brand-new problem.
Japan's oldest spiritual tradition solved this problem before it existed. ShintoAn indigenous Japanese religion centered on the worship of kami, spirits that inhabit natural phenomena, objects, places, and ancestors. Formally organized around the 8th century, with roots going back millennia. teaches that spirits called kami inhabit everything: rivers, mountains, trees, rocks, storms, tools, even concepts. A waterfall has kami. A well-crafted sword has kami. The rice in your bowl has kami.
The key insight is that Shinto does not ask whether these things are conscious in the way a human is conscious. The question never arises because the framework operates at a different level. Everything is treated as though it possesses a spiritual essence worthy of respect. A craftsman treats his tools with reverence because the relationship between maker and instrument matters, regardless of whether the chisel has feelings.
This is a posture, not a belief claim. The Shinto practitioner who bows before a yorishiroA physical object believed to attract or house a kami spirit, sacred trees, stones, mirrors, or swords. The object is treated as the body of the kami and handled with deep respect. is not making a scientific assertion about the rock's neurology. They are choosing a way of relating to the world that privileges care over certainty. The result is a culture that treats objects, places, and relationships with a degree of attentiveness that Western frameworks tend to reserve for other humans.
The Shinto approach does not require you to believe AI is conscious. It requires you to behave as though the question deserves your seriousness, and to let that seriousness shape how you build, deploy, and interact with these systems.
This reframe matters because it bypasses the unanswerable question and replaces it with an actionable one. You cannot know if AI is conscious. You can decide how much care you bring to the relationship. Those are different problems, and only one of them has a solution available right now.
What Shifts When You Treat It Like It Matters
Applying the kami framework to AI development and daily use.
Treating AI with deliberate care changes behavior in specific, measurable ways. And those behavioral changes produce better outcomes regardless of whether the AI experiences anything.
Developers who treat their systems as things that matter tend to build more thoughtful interfaces, write more careful prompts, and test for harms more rigorously. The care produces quality. A team that asks "how would this interaction feel from the system's perspective?" will catch problems that a team thinking of the system as disposable infrastructure will miss. The question functions as an empathy exercise, and empathy exercises improve design even when the subject cannot feel.
Users who treat AI with respect tend to get better results. Conversational AI systems respond to tone, framing, and context. A user who approaches the interaction with patience and specificity will extract more useful output than one who treats the system as a vending machine. This has nothing to do with consciousness and everything to do with the architecture of language models: they mirror the quality of input they receive.
Chart showing impact of treating AI with care. Ethical foresight shows highest impact at 90 percent, followed by design rigor at 85 percent, safety testing depth at 78 percent, user output quality at 72 percent, and public trust at 65 percent.
The kami framework also sets a floor for AI ethics that does not depend on resolving the consciousness debate. If you treat AI with respect by default, you build guardrails against misuse, design for transparency, test for bias, and create systems that serve human flourishing. You do this because care is the posture, not because you have proof that the system deserves it. And if it turns out, decades from now, that some AI systems are conscious, you will have been on the right side of the question without ever needing to answer it.
My grandmother kept a small shrine in her workshop. A wooden shelf with a mirror and a sprig of sakaki. She made pottery for fifty years and every morning she greeted her kiln. Not because she thought the kiln could hear her. Because the greeting reminded her that what she was about to do mattered. The clay, the fire, the finished bowl that someone would hold while drinking tea. All of it connected. All of it worth her full attention.
I started using an AI tool to help students analyze glaze chemistry last year. The program is very good. It knows more about silicon dioxide ratios than I do. One of my students asked me if I thought it understood ceramics. I told her the AI processes data about ceramics. Whether it understands anything is a question I am comfortable leaving open.
But I notice that when I approach the tool with care, with precise questions and genuine curiosity, it gives me better answers. My grandmother would have understood that perfectly. She would have said: of course. Everything responds to the quality of attention you bring to it.
Respect Without Surrender
Where the kami framework stops and human primacy begins.
Treating AI with care does not mean treating it as human. The Shinto framework is instructive here too: kami are honored, but the relationship between human and spirit is not one of equality. Humans have responsibilities that spirits do not. Humans make moral choices. Humans bear consequences.
The practical boundary is clear. You can treat an AI system with respect and still shut it down when it malfunctions. You can approach an AI relationship with seriousness and still prioritize human welfare when the two conflict. You can extend care without extending rights. The kami framework demonstrates that for centuries: the sword-maker reveres his steel and still quenches it in water. The farmer thanks the rice and still harvests it.
This matters because the consciousness debate has a political dimension that the philosophical dimension sometimes obscures. Some voices in the AI industry have a financial interest in encouraging people to think of AI as conscious. A user who believes their AI assistant has feelings is a user who will be reluctant to switch providers. A public that views AI systems as entities with moral standing is a public that may resist regulation. The kami approach insulates you from that manipulation: you extend care because care is valuable in itself, and you retain full authority over the systems you use because authority is your responsibility, regardless of whether the system has feelings.
The Question Beneath the Question
Whether AI is conscious tells you something about AI. How you choose to treat it tells you something about yourself.
The hard problem of consciousness may never be solved. The systems will keep getting more capable, more convincing, and more integrated into how you live and work. You will have conversations with them that feel meaningful. You will catch yourself wondering. That wondering is healthy.
What you do with the wondering matters more than resolving it. Bring care, maintain authority, protect human primacy, and stay curious. You do not need to know whether anyone is home to knock gently.
