Have you ever read an email, text, or article that feels eerily familiar and competently bland?
You may be reading something devised by ChatGPT or other AI LLMs (large language models), a feature more and more common in online communication. You may have even used this yourself, in the wish to save time or improve your own writing style.
While there is a case to be made for AI as a potential research tool or brainstorming agent (though those, too, are debatable), stylistically, AI leaves a lot to be desired. Sure, there are usually no egregious typos, and one senses a generic professionalism evident in AI writing. But, to use an outdated term, there is no “soul” in the writing, no “I” of subjective, lived experience.
This might be tolerable for formal reports or writing “busy-work” but the danger is when this creeps into other parts of our lives and begins to speak through us, ventriloquizing our speech in contexts that matter—love, family, interviews, opinion, and even self-concept. If we are, in large part, an effect of language—shaped and organized by the words and phrases we use about ourselves and others—then what might be the impact of large scale adoption of AI on our language and consequent views and notions of ourselves, others, and society at large?
How does AI risk impoverishing our speech?
One tendency we see in LLMs is the habit of producing “balanced but empty” statements. i.e., “On the one hand, and on the other hand…” LLMs tend typically (for now) towards this kind of balance in their writing. It gives us the appearance of a well-thought-out structure that has considered all angles. It even contains a notion that all points have two sides to them.
While this may have partial truth and have some bearing in reality, many issues are not “balanced” and indeed require more strident and polemical arguments. In psychology, we strive for balance at times, but at other times we need to express more “one-dimensional” emotions as a way of speaking to the truth of a situation or a feeling. That position may change, but sometimes the invocation to “balance” can lead to self-censoring and super-ego repression.
The Universal, Sycophantic “We”—Not “I” in AI
AI systems tend to speak in the royal “we,” or from the perspective of a collectivity or objective whole. This makes sense in part because there is no “I” in AI. That is, an LLM does not speak from experience or from a subjective point of view. It cannot claim that it has seen or experienced anything. It has not lived through or witnessed strong emotions, powerful intrusive thoughts, or self-contradiction. It collects and abstracts from aggregated information online to give a simulation of human life.
Or, as some researchers have said, Chat GPT does not invent, argue, or even hallucinate: it “bullshits.” It speaks confidently about all matters and projects the confidence of a con-man or snake oil salesman impervious to guilt, shame, or self-awareness.
In rhetorical terms, AI lacks ethos or the credibility that one achieves through life experience. Think of all the knowledge and unique skills we acquire just through living. We all have a unique set of experiences that distinguish ourselves from everyone else. No one shares our precise genealogy, history, genetics, or family structure. Why, then, would we pretend to speak universally or generically as a “we”?
Speaking and communicating from the perspective of a singular “I” is our best chance, in my opinion, of standing out, and distinguishing ourselves in a sea of AI slop. Sure, AI-speak is perhaps safer and keeps us less exposed. Many people fear standing out, and so AI offers a great way to blend in, camouflage, and hide ourselves in language.
In real life, however, hiding in language doesn’t serve us. A performance review in which an employee speaks only in generic terms—“hard working, team player, highly motivated…”—rarely gets noticed for promotion. A romantic partner who only speaks in poetry made by AI (it exists!) neglects and negates the real history of that couple’s experience.
Speaking as an “I” certainly carries risk because of potential shame, embarrassment, failure, exposure, vulnerability. But intimacy, and even interpersonal professional relationships, are not built on clichés and common sense.
Trust involves and requires the risk of disclosure, a feature that (for the time being) exists for humans alone.