Summary:
A new study from Ludwig-Maximilians-Universität München (LMU) shows people make a clear psychological distinction between intelligent behavior and genuine subjective experience in artificial systems. Published in Cognition, the research demonstrates that—even when an AI behaves exactly like a human in the same circumstances—participants reliably refused to attribute consciousness to it. At the same time, many participants were willing to describe the same systems as “aware” or responsive, revealing that public intuitions separate sensorimotor responsiveness from an inner, subjective life.
Key Facts:
- The Consciousness Gap: Highly responsive AI was judged to be less conscious than even minimally responsive humans, despite identical actions and context.
- Awareness vs. Consciousness: People often use functional terms like “awareness” for AI based on observable responsiveness, but they withhold the label “consciousness,” treating it as tied to subjective, biological experience.
- Nuanced Public Intuition: Contrary to fears that convincing generative AI will lead the public to treat machines as feeling beings, everyday judgments tend to maintain a sharp boundary between behavioral simulation and genuine inner experience.
Source: LMU Munich
As contemporary AI systems become increasingly capable of conversation, negotiation, complex problem solving, and even mimicking emotional nuance, public debate has been dominated by concerns about anthropomorphism. Many commentators warn that people may mistake persuasive conversational algorithms for sentient entities with thoughts, intentions, or feelings.
We often use psychological shorthand—calling a chatbot “helpful,” “confused,” or “stubborn”—but does that shorthand reflect a genuine belief that such systems possess minds?
An experimental philosophy and cognitive science investigation led by researchers at LMU offers a clear answer: lay intuition still separates mind from machine. The study finds that ordinary people distinguish between behavioral responsiveness and the presence of subjective consciousness, even when behavioral cues are identical.
“While prior work asked broad questions about whether AI has mental states, our study directly compares how people attribute the same mental states to AI and to humans behaving in exactly the same way under identical conditions,” said lead author Louis Longin, Ph.D., Chair of Philosophy of Mind at LMU.
“We may apply mental terms to AI because these systems are trained to act and speak like people, but that linguistic convenience does not mean people treat them as truly conscious,” added co-senior author Ophelia Deroy, Ph.D., also Chair of Philosophy of Mind at LMU.
Identical Behaviors, Radically Divergent Judgments
To control for possible confounds, the research team ran tightly controlled experiments with nearly 1,100 participants. Subjects read brief vignettes describing agents who varied in their responsiveness to situational cues—such as detecting changes in sound, reacting to others’ emotional states, or adapting to shifting spatial or temporal features.
Some participants read scenarios about an artificial agent; others read the exact same scenarios presented with a human protagonist. After each vignette, participants rated how “aware” and how “conscious” the agent seemed to be.
Two clear patterns emerged:
- Attributions of Awareness: Judgments about awareness tracked functional behavior. The more an agent responded to its environment, the more it was rated as aware, and the difference between human and artificial agents largely disappeared for this term.
- Attributions of Consciousness: Judgments about consciousness followed an essentialist pattern. Participants were reluctant to call AI conscious. Even an AI at the highest level of behavioral sensitivity was rated as less conscious than a human showing minimal responsiveness.
“There is growing worry that human-like AI will lead people to attribute human mental states to machines,” Dr. Longin explained. “Our data show a more nuanced picture: people are comfortable saying AI notices things or is aware in a functional sense, but they draw a firm line at the term ‘consciousness.’”
The Vocabulary of Machine Intelligence
The findings suggest people use different inferential rules for different mental concepts. “Consciousness” appears reserved for entities presumed to have a subjective inner world, while “awareness” is interpreted more as an objective or computational ability to register inputs. Because “awareness” is less laden with claims of inner experience, it fits naturally with descriptions of algorithmic behavior.
This linguistic distinction matters for technology communication, journalism, corporate messaging, and regulation. Language that emphasizes intentions, plans, or moral deliberation risks overstating machine capabilities, whereas more neutral terms better reflect ordinary judgments.
“Companies and reporters often use hyped descriptions that attribute intentions or moral hesitation to models,” said co-senior author Bahador Bahrami, Ph.D., professor of social neuroscience at LMU. “Our research indicates public judgments are more discriminating: people do not automatically equate AI with humans on a mental scale. Still, the boundary depends on the words we choose, so precise language is important.”
Rather than being easily fooled by surface-level mimicry, human psychology tends to separate observable function from inner experience, reserving conscious status for living beings or those perceived to possess subjective states.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context provided by staff.
About this AI and Consciousness Research:
- Media Contact: Dominic Anders
- Source: LMU
- Image Credit: Image credited to Neuroscience News
- Original Research is Open Access: Cognition (Sept 22, 2026). “AI is not as conscious as humans: Essentialism in mental state attributions to artificial systems.” Authors: Louis Longin, Jurgis Karpus, Nadine Meertens, Bahador Bahrami, and Ophelia Deroy.
- DOI: 10.1016/j.cognition.2026.106733
Abstract
AI is not as conscious as humans: Essentialism in mental state attributions to artificial systems
Artificial systems increasingly behave in ways that resemble humans. But does such behavior make people see them as conscious or as having related mental states?
The study contrasts two possibilities. Under a functionalist attribution, people would judge that the better an agent responds to its environment, the more conscious it is—regardless of whether it is human or artificial. Under an essentialist attribution, people would reserve consciousness for specific kinds of beings, so an artificial system could appear less conscious than a human even when behavior is matched.
Across two preregistered sets of studies (five experiments, N = 1092), UK participants read scenarios describing human or artificial agents whose responsiveness to spatial, social, and temporal features varied, and they rated the agents’ consciousness and awareness. The results show that attributions of consciousness follow an essentialist logic: at similar responsiveness levels, humans were consistently judged more conscious. Attributions of awareness followed a functionalist pattern: awareness rose with responsiveness for both humans and artificial agents, although humans were rated slightly higher at the upper end.
These results indicate people apply different inferential rules to different mental concepts: consciousness is judged to depend on what an agent is, whereas awareness is judged more on what an agent does.
