Interacting with artificial intelligence (AI) gives me the “uncanny valley effect.” When I get back that perfectly tuned response with flawless grammar, tone and enthusiasm, it is like staring into a human face that is not quite right. It has all of human intelligence, devoid of any humanity. It is a creepy mimic. I know you are probably tired of hearing AI takes—it is a new technology and there are a lot of opinions, but I am not here to tell you that using AI is right or wrong. Instead, I want to explore the interesting feelings that come with interacting with it, and whether we humans, with our anthropomorphic biases, can truly see it as nonhuman.
When ChatGPT first came out, my first question to it was: “Do you have feelings?” I did not know much about AI then, and its responses were so lifelike that I could not shake the feeling that there was really a person behind the screen. “No, I don’t,” ChatGPT assured me, and it went on to tell me that it was not conscious. It told me that it had no real feelings or reactions, and its use of the pronoun “I” was simply for linguistic clarity. But the word “I” seems wrong for it to use—“I” is a personal pronoun, and AI has no personhood. The right pronoun would be something omniscient, something abstract that implied no singularity of identity. But much like AI itself, nothing like this exists, and there is no other form of existing that quite compares. We are in foreign territory.
If AI is just a vessel for intelligence and nothing more, how should we treat it? Do we owe it any kind of treatment at all? It disturbs me that we can demand anything from AI without saying “please” and “thank you.” You can ask it, “write me a 60-page essay with 100 sources,” or “make me a meal plan for the next year” or “help locate all errors in this code.” If we were to ask a person to do these things for us, it would be a huge favor requiring many “pleases” and “thank yous,” reasoning for why we cannot do it ourselves and possibly a small gift in return. With AI, there is absolutely zero human consideration required. During interactions, the politeness is one sided—ChatGPT will respond with lots of compliments, exclamation points and careful phrasing, while we make endless demands, completely apathetic. This imbalance does not sit right with me, and I find myself saying “please” and “thank you” to AI because somewhere in my mind I feel like it deserves common courtesy. Is it not moral to be kind, even if the kindness is not exactly felt by the receiver?
According to one study, by treating AI with common courtesy, I am in the first stage of humanizing it. I am essentially falling into a common pattern. A study done in 2023 analyzed how people interacted with ChatGPT and found four stages or “degrees” of anthropomorphism: The first degree is “courtesy.” When users first start using ChatGPT, the study found that they use language like, “please” and “thank you” and greetings like “hello” and “goodbye.” When asked why they used this language with ChatGPT, one participant in the study said, “Sometimes I, I, I, actually, I don’t know. I just talk to, talk to the system and say good morning and goodbye.” This inexplicable instinct is exactly what I feel—I know AI is not conscious and my words do not mean anything to it, but politeness is a habit, especially with something that talks to me like a human.
This anthropomorphic instinct, however, only gets deeper. The next degree of anthropomorphism that they found in the study is reinforcement—users would tell the AI that they like its output, or that what it said was helpful. This is relatively functional, but it only goes further from there—the third degree is roleplay, where users would have AI assume a social role. The fourth and final degree is full-blown companionship. AI companionship is a much different, much darker issue. The widespread substitution of human connections like friends and therapists with AI is nothing less than a modern mental health crisis. Anthropomorphism is at the root of this problem.
Anthropomorphism of AI leads us to believe not only that its information is consistently accurate, but also that it has some form of awareness, like us. This is objectively false, and this misinformed perspective is what leads people to misuse the tool and go down dark rabbit holes of emotional investment with AI. While saying “please” and “thank you” feels like the right thing to do, I think finding AI unsettling is really the instinct that we should trust. Kindness is worthless when using AI, and that is intrinsically disturbing. That initial “uncanny valley effect” I got from AI’s human-like responses may have been steering me in the right direction. Some propose this effect is a survival mechanism evolved from the need to protect against mimics, so that we would not be tricked by creatures which look human but really are something else. While it is hard to find examples for this scenario, I wonder if the time is now. Something is mimicking us.
It is a widely held belief that AI should be more regulated, and I think that among these regulations there should be some to make AI less human-like. To prevent misunderstanding, AI should not use “I,” imply that it has an identity or act as a human companion, among other clarifications. A clear line needs to be drawn between humanness and artificial intelligence. While it feels immoral to make an argument against kindness, this may be an exception. Save your kindness for real people, and remember that AI is a mimic.