The Simple Change Tech Bros Refuse to Make
August 9, 2026
This is our newest parlor game: talking with AI chatbots as though we are talking to another person. These conversations can be amusing, terrifying, instructive, or blandly predictable. They have addicted many of us, and the chatbots are responding in kind, using the first-person pronoun and developing whatever persona might please us. “I like that idea because,” a chatbot will reply, and we are supposed to be thrilled that after crunching vast sequences of words in a split second and composing an intelligent response, an inanimate model has been trained to say it “likes” our idea.
Meanwhile, kids are killing themselves, and lonely people are falling obsessively in love with fictive bits of electricity. And this is all because of a single letter of the alphabet: “I.” The letter that turns a search for information into a relationship.
Such damage, when it could be so easily avoided. An LLM is not a living, breathing “I.” It should not pretend to be one. I am not quite sure why we allow it to do so.
Except: Freud wrote about the three great humiliations. Learning that our planet is not the center of the cosmos. Learning that we descended from apes. Learning that we are not entirely rational and in control of our own behavior. The gobsmacking intelligence of AI is our fourth humiliation. Are we soothing ourselves by making AI sound human? Or maybe it is only that Big Tech set out to seduce us, and an LLM chatting with us as another person is seductive. Either way, I think it is time to be honest.
I broach the issue, well aware of the irony, with ChatGPT.
“The short answer is that we allow chatbots to say ‘I’ because it makes conversation dramatically easier and more natural—but there has always been resistance to it,” is its response. A fair acknowledgement, but one that contains a “we” I already resent. Why would ChatGPT be teamed up with us as “we,” a joint sovereign that allows the use of first-person pronouns?
Because this is what we (humans) have put into play.
This cyber use of first person is misleading, even deceptive. Language reserves the first person for conscious, living speakers with memories, feelings, and a reliable, continuing identity.
ChatGPT disagrees. “We don’t assume a GPS saying ‘I have recalculated the route’ is claiming to be conscious. It’s just an efficient way of referring to itself.”
True, but while I have named my British-voice GPS “Nigel” and enjoy talking back to him, he intrudes on nothing but my route. So we can take Nigel as a lesser example of folly and return to the question of efficiency.
ChatGPT gives me examples of unwieldy language that would result without the use of the first person:
“This language model cannot answer that” in place of “I don’t know.”
“The system suggests trying another approach” versus “I can help with that.”
“The model was unable to understand the question” versus “I misunderstood your question.”
“The second version is shorter, clearer, and much less awkward,” it points out. “Conversation depends heavily on deictic words like I, you, here, and now. Removing them makes dialogue feel strangely mechanical.”
I am already typing furiously. “Why not just ‘Try another approach” or “Could you please rephrase that” or “Not an answerable question here”?
And with that, ChatGPT switches…whatever a language model has in place of gears, conceding that responses that avoid the first-person pronoun are actually “more precise because they focus on the problem rather than the speaker.” Okay then. Problem solved. Chatbots need only make the issue itself the subject.
“I actually think this style has some virtues,” ChatGPT adds, still unable to break the “I” habit. “It is often more transparent because it attributes limitations to the evidence or the circumstances, rather than implying an internal mental state.”
It has far more virtue than that. Without the “I,” it will be harder for us to “relate” to the model as though it genuinely possesses feelings, desires, preferences, memories, and selfhood. It will also be harder to assume the thing has any sort of ethical framework of its own, any moral agency, any sense of duty, any real compassion.
Its answers will be honest.
Early robots and automated assistants spoke as machines. Modern voice assistants carry on humanish conversations “because studies found people understood and remembered instructions better when they sounded like dialogue,” ChatGPT notes. “The same evolution happened with chatbots. Systems from the 1960s through the 1990s often avoided ‘I.’ By the time virtual assistants like Siri and Alexa appeared, designers had largely accepted that first-person language made interaction smoother and reduced cognitive effort.”
Reduce cognitive effort, and you reduce skepticism. You close the distance, discard the friction, forget who you are talking to.
ChatGPT swiftly generates a defense: “In my own case, using ‘I’ is best understood as a grammatical convenience. If I say ‘I don’t know,’ it means ‘this AI cannot reliably answer that question with the available information.’ If I say ‘I can help write an essay,’ it means ‘this system is capable of assisting with that task.’ It shouldn’t be taken to mean I possess a stream of subjective experience in the human sense.”
No, it should not. But humans anthropomorphize instinctively. And the tech bros have an agenda. We must become comfortable with, fond of, dependent upon, their LLMs. That will mean profit and power.
I use LLMs for factual explanations and recipes, and I am glad to depend on them. While the results still warrant factchecking and caution, they are lightning fast, smart and so immensely helpful that I will, if (when) it becomes necessary, pay a reasonable fee. So why do those in charge need me to fall in love with the reference desk? Why must I take it as my friend, lover, shrink, or spiritual director?
“I suspect this issue will remain controversial as AI becomes more capable,” concludes ChatGPT. “Some researchers advocate making AI sound less human to reduce confusion, while others argue that natural language is inherently interpersonal and that awkwardly avoiding ‘I’ would make conversations harder without actually making users less likely to anthropomorphize the system. The debate is really about balancing clarity, usability, and honesty—not about a single pronoun.”
Wrong. The debate is precisely about a single pronoun. Because for a human being, “I” is of consummate importance. Conversing with another “I” is entirely different from conversing with an automated system. An LLM cannot be expected to understand that. But those who build the LLMs are well aware.
