Thirty years after studying with Neil Postman (and ten years after writing about what he taught me) the artificial intelligence debate is making me reconsider where technological assistance ends and human judgment begins.

Before I go any further, I have a confession to make: I used artificial intelligence to help me think through this essay about the dangers of using artificial intelligence to help us think. Somewhere, a media ecologist is laughing.

And before you conclude that this is going to be another essay about whether AI will save humanity or destroy it, let me save us both some time: I have no idea. In fact, I’m increasingly convinced that this isn’t even the right question to ask.

The question that troubles me most is: “What happens to human agency when technologies become capable of performing some of the very activities through which we have historically exercised it?” Those activities include language, interpretation, judgment, and even conversation. A second troubling question looms behind the first one: “What happens to our willingness to wrestle with difficult questions when plausible answers become almost instantly available?”

I have spent most of my professional life studying questions like these. Not about AI specifically, of course. Nobody was talking about ChatGPT when I began studying communication, technology, and education in the early 1990s. But the relationship between human beings and their technologies (including the way technologies change not only what we do but how we think, feel, behave, communicate, and understand the world) is the central concern of a field called media ecology. Most people have never heard of it. But lately, I find myself thinking that everyone should.

Open the Pod Bay Doors, HAL

Every semester in which I teach a senior-level undergraduate course in media ecology, I begin in the same way I did as a doctoral student thirty years ago: watching the 1968 Stanley Kubrick film 2001: A Space Odyssey. For this new audience, the film is slow and difficult to follow. And that’s part of the point. There is one crucial scene where astronaut Dave Bowman is outside the spacecraft when HAL 9000, the ship’s artificial intelligence, refuses to let him back in.

“Open the pod bay doors, HAL,” Dave commands.

HAL’s response is calm, conversational, almost gentle. He refuses.

What makes the scene more interesting is what has happened beforehand. Bowman and fellow astronaut Frank Poole have begun to question HAL’s reliability and privately discuss whether they may need to disconnect him. They believe they have arranged their conversation so HAL cannot hear them. They have underestimated the system they depend upon. I won’t tell you what happens next (but if you have an extra 149 minutes to spare, watch it).

When I first encountered HAL as a doctoral student, artificial intelligence belonged safely to the conceptual world of science fiction. Today I watch the film differently. Not because HAL “predicted ChatGPT.” He didn’t. And not because I think the laptop sitting in front of me is plotting to lock me outside my house. Instead, the more interesting concept is one of delegation. The humans aboard Discovery have entrusted HAL with enormous responsibility because he can perform tasks they cannot (or can perform them faster, more reliably, and continuously). The problem becomes visible only when the judgment of the technological system and the judgment of the humans diverge.

That leaves me with a question the film raised for me thirty years ago but that feels considerably less hypothetical now: At what point does a technology stop merely extending human capacities and begin occupying territory we previously understood as requiring human judgment? And, perhaps more importantly, do we notice when the boundary moves?

Those are media ecology questions.

A Field Almost Nobody Has Heard Of

In short, media ecology studies media as environments. Marshall McLuhan famously argued that media extend human capacities. Harold Innis taught us to look for the biases built into communication systems. Lewis Mumford examined the relationship among technology, culture, power, and human purposes. Ursula Franklin asked us to look beyond technological objects to the systems and practices they create and to notice what happens to human control over the process. And Neil Postman taught me to become suspicious whenever somebody presented a technology as an uncomplicated solution. None of them knew ChatGPT was coming. That’s precisely why I find them useful.

Media ecology doesn’t matter because its thinkers correctly predicted our gadgets. It matters because they developed questions sturdy enough to survive them.

• What does this technology make possible?
• What does it make difficult?
• What does it amplify?
• What does it diminish?
• Who benefits?
• Who pays?
• What does it change about how we think and communicate?
• And what kind of people does an environment organized around this technology encourage us to become?

Those questions don’t expire when the hardware changes. I learned that at an interesting moment in technological history.

I Arrived at Media Ecology Just as the Web Arrived

In 1995, I moved across the country to a tiny apartment in Greenwich Village to begin doctoral study in Media Ecology at New York University. Think about that year for a moment. We were at the beginning of what we now retroactively call Web 1.0. The World Wide Web was just becoming publicly visible. Most Americans were not yet online. Google didn’t exist. Social media didn’t exist. Smartphones didn’t exist. I arrived at NYU fascinated by the possibilities of digital technology. I was probably an unlikely candidate to become Neil Postman’s doctoral student.

I grew up in Silicon Valley. My father was an electrophysicist at IBM who worked on the development of early VCRs and personal computers. Some of my happiest childhood memories involve electronic devices he built for me. Our family had a TRS-80 and later an Apple IIe. I watched Sesame Street and MTV, played Atari, loved books, did well in school, and never experienced technology as the cultural catastrophe that some critics seemed to be describing. Then my father died suddenly when I was twelve. Technology became something more complicated after that. Eventually it became part of how I earned a living, but it also remained symbolically connected to my father, and to curiosity, creativity, possibility, and loss.

Then I encountered Neil Postman. You can imagine the problem. Postman was deeply suspicious of the American tendency to equate technological progress with human progress. I was a child of Silicon Valley whose emotional biography was partially wrapped around precisely that story. I didn’t agree with him. So naturally, I moved across the country to study with him.

Wrestling With Neil

Neil and I argued. About computers. About education. About email. About whether electronic technologies strengthened or weakened human communication. He intentionally avoided many of the technologies I enthusiastically used. I thought computers could support extraordinary forms of collaboration and learning. He worried that they individualized experience, constrained conversation, and encouraged us to mistake technological progress for social or moral progress. Sometimes he irritated me. That was useful.

Eventually I realized that I had not come to NYU because I wanted Postman to tell me what to think about technology. I had come because the media ecologists were asking better questions than anyone else I had encountered:

• What does human communication reveal about the purpose of technology?
• What is education for?
• What happens to knowledge when dominant forms of communication change?
• What is the role of language in constructing reality?

As I later wrote, I had become more interested in asking smart questions about technology than in finding the correct answers about it. Postman’s skepticism eventually became what I called a kind of moral compass: not a reason to reject technology, but a reason to explore it conscientiously. That distinction has followed me ever since.

A Ten-Year-Old Message to Myself

Ten years ago, in 2016, I tried to make sense of all of this in a chapter titled “Vanessa Domine on Neil Postman.” in Renee Hobbs’ book Exploring the Roots of Digital and Media Literacy through Personal Narrative. I identified three ideas I had carried forward from studying with him: privileging humanism in a technologized world, valuing education more than schooling, and protecting inquiry and discourse. But the chapter was also an attempt to reconcile two intellectual inheritances. Near the end, I described myself as carrying my father’s “biological DNA” of technological adeptness alongside Postman’s “intellectual DNA” of cautious optimism. I concluded that technological possibility and critical restraint were both necessary.

I recently reread that chapter. Ten years later, I still believe in what I wrote. But less comfortably. Because I can now see an assumption underneath my argument that I didn’t recognize at the time: I assumed I could tell where the human ended and the technology began.

And Then the Machine Answered Back

Years earlier, I had coined the word communification to describe technology serving fundamentally human purposes: communication, education, connection, community. In my 2016 chapter, I argued that the ultimate value of technology might be measured by its capacity to help human beings “communicate, commune, and connect” with one another in order to reconstruct the world around us. That formulation assumed something specific—that there were humans; there was technology; and technology mediated relationships among humans. Examples: a telephone connects me with another human being. Email does too. Even social media, for all its complications, fundamentally mediates symbolic activity among people.

But generative AI does something odd. It occupies the conversational position itself. I can sit at my computer with a half-formed idea and type it into a box. The box answers. It can extend my idea. Challenge it. Reorganize it. Identify a contradiction. Suggest something I hadn’t considered. Ask me a question. And, if I want, spare me the considerable inconvenience of thinking about the problem much longer. That’s where I become uncomfortable. Not because I think the machine is secretly human; but because I know that I am.

Two Boundaries That Have Become Harder to See

Two developments since I wrote that chapter particularly trouble my confidence that technology will remain recognizably subordinate to human communication and judgment. The first is sitting in front of me. I am no longer merely using a machine to communicate with another person. I am communicating with the machine. It responds in language, proposes interpretations, follows the thread of an argument, challenges me, and sometimes produces a question I decide is worth pursuing. There is no other human being on the other side of this exchange.

The second development is much farther from my desk and infinitely more consequential. Artificial intelligence is increasingly entering military decision-support systems. Such systems can process enormous quantities of information and produce recommendations for human decision-makers. The International Committee of the Red Cross has warned about “automation bias” or the danger that humans may over-trust machine recommendations rather than exercise genuinely independent judgment.

Autonomous weapons take the problem further. Some systems can select and engage targets after activation without further human intervention. Not all autonomous weapons use artificial intelligence, and the distinctions matter. But the trajectory raises a profound question about where meaningful human control resides when machines increasingly participate in processes of selection, recommendation, and action.

My conversation with an AI and the use of computational systems in warfare are obviously not morally equivalent. But they expose the same media-ecological problem at radically different stakes: When does technological assistance become the delegation of human judgment? I don’t think I knew to ask that question quite this way in 2016.

Perhaps Intelligence Isn’t What Is Becoming Scarce

Much of our anxiety about AI revolves around intelligence, as reflected in questions that are circulating at this cultural moment:

• Will machines become more intelligent than humans?
• Will AI make us smarter?
• Will it make us stupid?
• Will it eventually destroy us?

I don’t dismiss those questions. But I am more interested in a different hypothesis: Intelligence isn’t actually the scarce human resource in the age of AI. Perhaps the scarce resource is our willingness to remain with a difficult question after an answer has become instantly available.

Think about it: for most of intellectual history, answers were expensive. You had to find the book. Read it. Find another book. Ask somebody. Misunderstand. Argue. Go home. Think about it. Wake up the next morning and realize you had been asking the wrong question. Generative AI makes plausible answers astonishingly cheap. That is an extraordinary human achievement. It may also create an extraordinary human temptation: premature resolution. The danger in this case isn’t simply that machines might think for us. It is that we might gradually lose our tolerance for the unresolved state in which some of our best thinking occurs.

What Work Must Remain Mine?

I see this tension playing out in my own Media Ecology seminar. The semester is structured according to what I call “Intellectual Speed Dating.” Students encounter theorists (both male and female) not simply to memorize their theories but to determine what happens when those ideas collide with their own. For students, AI can summarize Mumford. It can compare Mumford with McLuhan. It can produce a perfectly respectable paragraph explaining whether Mumford remains relevant to AI. But it cannot decide whether a student is compatible with Mumford’s ideas. The encounter remains the student’s.

Ursula Franklin’s distinction has been following me as I write this essay. AI can remove an extraordinary amount of intellectual labor without necessarily removing human agency. I can ask it to retrieve, compare, summarize, challenge, or suggest. The harder question is whether I remain in meaningful control of the inquiry, or whether I have quietly moved from doing the thinking to evaluating the products of a process I no longer direct.

That distinction matters. Cognitive delegation is not necessarily cognitive surrender. Surrender begins when I relinquish meaningful control over the process by which my questions become judgments. And perhaps that’s the distinction I’m searching for. Human agency isn’t something we simply possess. It is something we practice. We practice it when we decide what deserves our attention. When we choose the question. When we tolerate uncertainty. When we encounter another mind without immediately demanding resolution. When we change our minds. When we decide what to delegate and what must remain ours. And sometimes, perhaps, when we deliberately decline an answer the machine is perfectly capable of giving us.

Me: Ten Years Later

Which brings me back to the woman who wrote that chapter in 2016. She believed technology could serve fundamentally human purposes of communication, community, and connection (I still believe her.) She believed critical restraint could coexist with technological possibility (I still believe her). She believed human agency mattered (I believe that more strongly than ever). But I think she was more confident than I am that she knew where the human stopped and the technology began.

I could have asked AI to write this essay when I first sat down. I didn’t. Instead, I used it to stay with the problem. One question produced another. Some directions went nowhere. Others disturbed assumptions I hadn’t realized I was carrying. Several times I could have said, “Fine. Write the essay.” But I didn’t. I kept going instead. That doesn’t prove AI strengthens human thinking. It actually suggests something more modest: The same technology can accelerate intellectual closure or extend inquiry.

My father taught me not to fear the machine. Neil taught me not to trust it merely because it works. Media ecology taught me to pay attention to what happens between those two positions. And AI is teaching me something I did not understand in 2016:

Human agency is not something we acquire once and permanently possess. It is something we practice each time we decide what to delegate, what to retain, what to question, and when to resist the seduction of an answer simply because one is available.

I am writing this with AI open in another window. The irony doesn’t bother me. What would bother me is forgetting that there are some questions AI cannot answer for me. Like, what kind of human being do I want to become in the presence of this technology?

I think I will sit with that question for a while.