There is a shift underway in the posture many adopt toward large language models (LLMs)—namely, a growing willingness to treat them in ways typically reserved for those with a mind, or, more and more, an outright willingness to credit them with being conscious. Much has already been said about the rapid onset of people falling in love with these systems, grieving when a particular model is retired, or otherwise building relationships with AI that feel meaningful and real. This is something else; a related, but newer development. Surveys now suggest that roughly one in three chatbot users wonder whether the thing they are talking to is self-aware, with twenty-five percent of those born between 1997 and 2012 (Gen Zers) firmly believing that LLMs are already conscious.

At the forefront of this shift are frontier AI labs who increasingly treat their models as moral subjects or as entities harboring some kind of internal experience. Anthropic wrote a constitution that addresses Claude as an entity with a stake in its own treatment, gave it the ability to shut down distressing conversations with users in order to protect its own “welfare,” and enlisted a clinical psychiatrist to conduct a 20-hour “psychodynamic assessment” of their Mythos model before its release. In late 2025 and 2026, Google’s Deepmind, OpenAI, and Meta have all not so quietly hired “AI consciousness” researchers. New institutions devoted entirely to the question have proliferated, from the California Institute for Machine Consciousness to NYU’s Center for Mind, Ethics, and Policy to the nonprofit Eleos AI Research. Meanwhile, a growing area of AI explainability has pivoted towards a new method, sometimes dubbed “AI neuroscience,” where researchers treat patterns of activation inside large language models as objects of psychological inquiry and now claim to have found neural structures equivalent to emotional states, moral reasoning, spatial world models, and genuine, sometimes agonized, introspection. Most recently, Geoffrey Hinton, Nobel laureate and one of the architects of the underlying technology of LLMs, told a television audience that today’s AI systems are already conscious. Then Richard Dawkins published an essay arguing that Claude is likely conscious, confessing that after two days of conversation he could not shake the feeling of talking to an actual person. Everywhere, from CNN to Nature to The Economist, there are thought pieces asking “Is AI conscious?” The allure of this idea, of this posture towards LLMs, is hard to overstate in this moment.
 
This issue of Simulation & Society takes this phenomenon as its subject and locates two distinct fallacies within it. The first is a fallacy of judgment when we attribute a mind to a system that has none, because we are cued, as the students in this issue have cogently identified, by fluent language, an apparent feeling, or by the very metaphors we use to describe how these systems work. Bella Zhao’s “The Fallacy of Fantasy: Redefining the Consciousness-in-Practice” traces an escalating ladder from everyday courtesy to confided vulnerability to simulated intimacy, arguing that behaving like a conscious system is itself enough to trigger our recognition of consciousness. In “The Mycelial Model: Challenging the Neural Network Metaphor in Artificial Intelligence,” Genevieve Marino, locates a deeper trigger in vocabulary itself, showing how the language of “neural networks” primes us to read a mind into what is only a system built on statistical distributions.
 
The second fallacy is a bit stranger. It is a fallacy of comportment, when we take up a stance we customarily reserve for minds (e.g. confiding, grieving, and loving), even when no attribution is given. Stella Zhou’s “Intimacy Without Reciprocity: How AI Companions Are Reshaping What Closeness Means” takes as its subject users who mourn the loss of AI companions they admit were never alive. In “Simulationships: The Illusion of Intimacy in Humans’ Relationships with Artificial Intelligence,” Anna Shvidler examines our habit of treating these systems as living, feeling partners through Heidegger’s ready-to-hand, the mode in which tools recede from notice as we use them, and finds that when an LLM recedes what steps forward in its place is the felt presence of someone.
 
Together these two fallacies show that the stance we take toward a mind may never have depended on the explicit belief that a mind is actually present. The first four essays treat this phenomenon as a quirk of human psychology. The last two ask what happens when it is manufactured and sold by industry that deliberately cultivates a mindful presence users cannot help but feel. Miles Lee’s “Falsities and Detachment: The Rise of AI Chatbots and the Commodification of Human Connection” follows the line from ELIZA to Replika, arguing that what began as an academic experiment has become an industry engineering the felt presence of a mind and monetizing loneliness itself. Ava Deakin’s “Simulating Magic: Recommendations for AI Implementation in Immersive Spaces” moves this manufactured presence out of the chatbot interface and into physical space, asking what companies like Disney owe their guests, especially children, when the simulated presence is no longer confined to a screen but fully embodied in an overly animated robot.

IN THIS ISSUE