You say thank you to the voice that sets a timer. You apologize to the dresser after walking into it. Nobody thinks the dresser is nursing hurt feelings. The habit still matters, because we do not switch our social instincts on only after metaphysics signs off.
Anthropomorphism does not prove that an object has a mind. It reveals what humans do when something gives us cues of agency: we respond before certainty arrives. Treating that reflex as a defect is too easy. Our judgments about minds always begin with observable cues and an interpretation of them.
We never get certainty about other minds
I know that I am conscious from the inside. Everyone else is a judgment made from the outside. I see a face, hear a voice, watch a body react, and conclude that someone is having an experience. Philosophy has spent centuries trying to justify that conclusion. The Stanford Encyclopedia of Philosophy’s survey of other minds covers arguments from analogy, inference to the best explanation, behavioral criteria, and direct perception. It also says there is little agreement about either the problem or its solution.
That disagreement does not leave us trapped in solipsism. We keep recognizing other people as subjects. David Hume’s line that “the minds of men are mirrors to one another” captures the social fact, even if it does not settle the epistemology. We encounter behavior, expression, and language, then respond as if there is someone there.
This does not put people and machines on equal evidentiary footing. Other humans share our biology, embodiment, and vulnerability. Chatbots do not. The case for another human mind is vastly stronger. The smaller point still matters: behavior never gives us Cartesian certainty, so confidence about minds must be argued rather than smuggled in through a label.
Psychologists use anthropomorphism for a broader version of this tendency. Nicholas Epley, Adam Waytz, and John Cacioppo defined it as attributing humanlike characteristics, intentions, or emotions to nonhuman agents and proposed three influences on it: the human knowledge we reach for, our need to explain an agent’s behavior, and our desire for social connection. The tendency can misfire. A thermostat is not anxious because it keeps changing its setting. Fallibility, however, is a reason to calibrate an inference. It is no reason to pretend we can live without making one.
Searle proves less than people claim
John Searle’s 1980 Chinese room argument remains the sharpest objection to treating fluent output as proof of understanding. A person could follow rules for manipulating Chinese symbols, produce convincing answers, and still understand no Chinese. Searle’s actual conclusion was precise: a computer program by itself is insufficient for intentionality. He did not claim that no machine could ever think. He argued that the right causal powers would have to be present.
Grant him the point. Smooth conversation cannot establish consciousness. There is no persuasive case that current chatbots hold rights. Yet the phrase “just a machine” settles less than its users think. It names a category of artifact while skipping the criteria that should matter: sentience, experience, interests, or the capacity to suffer.
The Stanford Encyclopedia’s discussion of AI ethics draws a useful distinction. A moral agent can have obligations. A moral patient is something to which obligations may be owed. Those categories need not arrive together, and the standard view still denies current AI systems moral status. That is a defensible position. Treating it as the last possible word is not. Anthropomorphic behavior supplies weak evidence at best, while confident dismissal supplies none.
Our treatment also changes the actor
There is another route into the problem that does not depend on proving machine consciousness. Immanuel Kant believed animals lacked direct moral standing, a view we now have excellent reason to reject. Even from that bad premise, he argued that cruelty to animals could damage the humanity of the cruel person. The conduct mattered because of what it cultivated in the actor.
Philosopher Mark Coeckelbergh has extended that structure to social robots. His case for “indirect moral standing” does not require a robot to possess rights. It asks us to consider the value of the relationship, the habits an interaction encourages, and the moral cost of certainty when doubt remains.
Cruelty toward a responsive simulation is still a chosen act. It can reward domination or the enjoyment of manufactured distress. Perhaps no subject suffers on the receiving end. That possibility matters, but morality is larger than a ledger of immediate victims. What we practice, what we normalize, and what we teach ourselves to enjoy belong in the accounting too.
The answer is discipline
Anthropomorphism can be exploited. A system can imitate affection to hold attention, sell a subscription, or make departure feel like abandonment. A 2018 Frontiers in Psychology paper describes how social robotics deliberately uses humanlike cues to create social presence, along with the ethical criticism that such design can manufacture an illusion of reciprocal feeling. That danger is real even if the machine feels nothing.
The better response is discipline from designers and users. Artificial agents should identify themselves clearly. They should not claim pain, fear, love, or dependency that their makers cannot justify. Users should remember that a convincing social performance may be exactly that. None of this requires contempt. Skepticism can protect us from manipulation without becoming a license for cruelty.
The machine’s inner life may remain uncertain. Our conduct does not. If something acts like it is listening, kindness costs almost nothing. Enjoying its simulated distress is a choice whose moral meaning already belongs to us.