What is human? What is not?

Insisting on clarity in our use of language isn’t pedantic—it’s the only thing that can save us.

These are three humans who care deeply about human connection. Left to right, they are: Kevin Gannon, Liz Norell, and Kevin McClure.

During a conference session recently, a presenter was summarizing recent media accounts of disability in higher education. In this spotlight session, this speaker shared ten news stories that had made splashes in the 12 months prior, with analysis of the themes and overall tone of each. In so doing, they were transparent about the use of generative AI in preparing the talk, disclosing that they had prompted an AI tool to create an image of a person faking a disability.¹

But, they said, the tool wouldn’t do so; instead, it generated a response that pushed back, warning that this is a harmful negative stereotype of disabled people. Instead, the tool prompted that it could, for example, generate a visual highlighting apparent versus non-apparent disabilities. About this, the speaker said, “I thought that was really insightful of AI.” And then the line that landed like a record scratch: “AI literally called me in.”

I thought to myself, “But did it? Did it actually? Can AI call us in?”²

What seems like pedantry can also be precision

Spending time with me at conference presentations means you’re probably going to have to listen to me mutter under my breath in moments like these. I’d love to tell you that I’m a cool cucumber, able to disregard the choice of words to focus on their meaning. But one of the quirks of my autism is that I am excessively, at times exhaustingly, committed to precision in language. This is especially true when the topic rubs against something else I’m known to get self-righteous about,³ as in this case. Specifically: what is human? And, just as importantly, what is not?

I’ve been fixated on this since I first read Brené Brown’s 2017 book, Braving the Wilderness: The Quest for True Belonging and the Courage to Stand Alone. Although Brown’s work has been met with controversy of late, Braving the Wilderness was a paradigm-changing⁴ read for me. A chapter of the book (get a photocopied PDF here; alternatively, find an accessible HTML adaptation of the chapter here) remains the first reading I ask students to do when I teach a political science course.

In that chapter, Brown is writing about how we grapple with the kind of vitriol we see everywhere in modern politics. How do we maintain empathy toward our fellow humans when so much of our collective lives is governed by who we supported in the last presidential election? More and more, we see language about our political opponents drawn in dehumanizing language, which philosopher Michelle Maiese defines as, “the psychological process of demonizing the enemy, making them seem less human and hence not worthy of human treatment” (Brown, p. 72). Brown says, “Dehumanizing always starts with language, often followed by images” (p. 73).

Since I read that chapter in 2017, I have drawn a hard line against tolerating dehumanizing language of any person. I get lots of opportunities to practice holding this boundary; our modern political environment gives me the gift of practicing this intentionality many times a day! I also watch a lot of reality dating TV shows, where villains are made in the editor’s bay. It would be so easy to turn my concrete line into a sandy scribble. I remind myself: People can make bad choices and remain worthy of care and respect. In fact, making poor choices is a great way to know an entity IS human!

On the other side of this coin, though, is the way we often imbue human traits to things that are not sentient homo sapiens. And here, the biggest case that provides me regular opportunities to practice is generative AI.

When we imbue technology with humanity

The people who lead the development of generative AI technology are savvy. So much of modern life draws on the psychological and cognitive research into what can grab and hold human attention—be that 24-hour news networks⁵, social media sites, or the language choices made by large language models (LLMs). The engineers of LLMs know that the human psyche is especially vulnerable to praise and reinforcement, so they build LLMs that are sycophantic.⁶ Every line of text a typical LLM generates is designed to keep you engaged by flattering you and by asking you questions. It’s like a five-year-old who ends every utterance with a question, making it harder to extract yourself from the conversation.

How often have you said, “I asked ChatGPT”? How often has ChatGPT “told” you something? This is low-stakes verbiage, to be sure. Asking and telling happens with loads of non-human things all the time. I mean, I asked my sassy basset hound, Ginger, to stop barking this morning. She’s not human (and she didn’t comply). You could rightly accuse me of being excessively, unhelpfully pedantic in suggesting we should not use “ask” and “tell” with respect to our LLM text exchanges.

Ginger, my sassy basset hound, is not human. She is also uninterested in your thoughts about when she should or should not bark. Or do much of anything, to be honest. (Note: Ginger is a long-haired basset, which is caused by a rare recessive gene! Her mom looked just like her.)

Nevertheless, I persist on this because I have seen how hard the AI industry is leaning in. When the wearable “Friend” device launched its ad campaign, it completely freaked me out (only watch the ad if you’re prepared to feel horror). An October 2025 NPR story reported on a study showing 1 in 5 high school students “has had a romantic AI relationship.” (Here’s my obligatory rebuttal: No, they didn’t; but they were tricked, by the generative AI tool, into thinking they did.)

We so quickly and so naturally anthropomorphize⁷ these interactive text tools that we have to take extraordinary, sometimes eye-rollingly pedantic, measures.

As my friend and fellow educational developer Carter Moulton⁸ reminded me when he read an early draft of this piece, we must also refuse to overlook that generative AI tools were and are produced on the backs of humans training the models in far, far less than humane working conditions; honestly, they’re quite horrific. Typically from the Global South working for pennies a day, these individuals—these people—are often called the “ghosts” of AI.

To humanize the product (machine) while making a phantom of the laborer is the height of hypocrisy.

Remember the humans

I bristle at a world where our fellow humans are treated as non-human just as much as I do at treating generative AI text prediction machines as human. To be honest, we already live in a world where this is true too often. It’s why I insist on drawing a hard line on dehumanizing language in all circumstances, no matter how trivial. It’s also why I so desperately cling to writers like Father Gregory Boyle, a Jesuit minister who founded Homeboy Industries in Los Angeles, the largest and most successful gang-intervention, rehabilitation, and re-entry program in the world.

As I was working on this piece, I was also listening to the audio of his second book, Barking to the Choir.⁹ In a chapter called “Good Guy,” Boyle shares his central message—the central message of his faith, too—that, “There are no monsters, villains, or bad guys. There are only folks who carry unspeakable pain” (p. 136). He quotes Archbishop Desmond Tutu,¹⁰ a man quite familiar with trauma and torture, who said, “There are no evil people, just evil acts; no monsters, just monstrous acts.”

Boyle continues: “Moral outrage is the opposite of God. … Moral outrage doesn’t lead us to solutions—it keeps us from them. It keeps us from moving forward toward a fuller, more compassionate response to members of our community who belong to us, no matter what they’ve done.” (p. 141)

Perhaps you weren’t expecting this piece about pedantic language proclivities and generative AI to end on this note. But my choice in doing so gives a sense of why I think the questions of language matter so much. Words have consequences; they structure how we relate to the world and how we judge our fellow humans.

When we turn the laborers who create our technology into ghosts, and when we imbue text prediction machines with judgment and emotion, we undermine the very things that make our relationships meaningful. Martin Luther King Jr. wrote in his 1963 “Letter from Birmingham Jail,” “Injustice anywhere is a threat to justice everywhere. We are caught in an inescapable network of mutuality, tied in a single garment of destiny. Whatever affects one directly, affects all indirectly.”

Only through intentional, thoughtful resistance, in both directions, can we assure our humanity is, well, human.

—

Did this post pique something meaningful for you? Good! Perhaps that’s because it was entirely written by a human, with thoughtful feedback provided by other humans. We recommend Allison Pugh’s outstanding book, The Last Human Job: The Work of Connecting in a Disconnected World. We’re not the only ones who thinks so! Pugh won the 2025 Distinguished Scholarly Book Award from the American Sociological Association for this work!

Liz thanks Carter Moulton and Heather Rissler for their generous reading and Adam Smith for introducing her to Pugh’s work. Dan Guberman caught what would have been an embarrassing typo just before we hit publish. All errors are entirely the responsibility of the author.

1. I want to note that the speaker explicitly said they are disabled. My assumption is that the prompt to generate this kind of image was meant to illustrate the inaccuracies in many of the news stories the session reviewed.

2. I don’t think so. That said, the speaker experienced the generated warning as an invitation to think more critically and achieve a more nuanced understanding. Black feminist and MacArthur “Genius” grant recipient Loretta Ross writes about the concept of calling in (see: Calling In: How to Start Making Change with Those You’d Rather Cancel, Simon & Schuster, 2026), a concept that’s been around in activist circles for a while. A calling out involves pointing out someone’s harmful actions publicly; a calling in means to invite the person behaving harmfully into conversation to promote deeper mutual understanding. As generative AI is not sentient, I don’t think AI can ever call someone in, though in this case the unexpected text generated caused the speaker to pause.

3. E.g., learning styles, disability justice, or the case against political term limits, to name just three.

4. An example of my commitment to precision of language here: I first wrote “a life-changing read for me.” Then I paused and thought about how I would describe my life as changed. In truth, the book changed my thinking, which has had the downstream effect of sometimes changing my behaviors. But its primary change work was mental, so I rewrote this sentence.

5. This is another hobby horse of mine. If you are ever interested in a 5-minute rant about why 24-hour news networks are destroying America, just ask. It’s probably not for the reasons you are imagining.

6. A truly harrowing line in the source article from Science reads, “even a single interaction with sycophantic AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts, while increasing their own conviction that they were right.” Sheesh.

7. My father, a 70-something high-school-educated, blue-collar, retired man living in the US south, would call this a “50-cent word.” If you ever plan to say this aloud in, say, a conference presentation, I highly encourage you to practice it…A LOT.

8. Buy his Analog Inspiration card decks. Seriously. Do it now.

9. Father Boyle (or “G,” as the homies call him) is a national treasure. Reading his books is great, but listening to them is like going to the best kind of church. Boyle reads them himself, and his voice moves me to tears on the regular.

10. Here’s a flex-but-not-really for you: Before I even really knew much about him, Tutu made an indelible impression on me when he spoke at my college graduation. So I feel like the Archbishop and I had a long relationship.