Where Are We Driving?

Spurred on by Hua Hsu’s alarming New Yorker article “What Happens After A.I. Destroys College Writing?,” the University of Puget Sound convened a pedagogical summit of sorts in October 2025 to take up the question “Is the big research paper dead?” Speaking for the history department, I said that it wasn’t dead yet—not unlike the plucky knight in Monty Python and the Holy Grail who keeps losing limbs but soldiers on. We have clung, through all these years, to the senior thesis—expecting our students to complete a major research paper in their capstone course as part of the history major.

The 1972 caption to this photo was “Children play in yard of Ruston home, while Tacoma smelter stack showers area with arsenic and lead residue.” Is AI now showering our students with fallout? Gene Daniels/US National Archives and Records Administration, Records of the Environmental Protection Agency, 545246/public domain

On the first day of that class—before I ask them to define history and reflect on why people find it interesting or useful—I start with an icebreaker: If you could have a historical superpower, what would it be? Some choose the ability to commune with the dead. Others would like to know what historical actors were really thinking when they did X, Y, or Z. Some want language fluencies. One student asked to magically be able to format citations in Chicago style, while others go big, asking for time travel. But no one has ever said about history research, “I’d like to just be able to push a button and be done with it.”

Writing a 25-page paper over the course of a semester is a journey. And if you ask me whether we should keep on doing what we’ve basically been doing with this capstone for decades, my answer might depend a lot on where we are in that journey. There’s plodding in the valley of doubt, and then climbing out. At so many points, I feel a kind of vicarious vertigo: Are we going to fall off this ridge and plunge into the abyss? Will this student finally find sufficient primary sources? Will another student engage in the relevant historiography? Will a third, bursting with ideas and passion for their project, ever put proverbial pen to paper?

But after the final drafts are in, I usually feel proud. The students forged on, in a community of learning. It’s not just a journey for a journey’s sake. It involves putting skills and hard work together, over the arc of a semester, into projects they care about, where they do the research, consult the experts, become experts themselves, argue their positions, work over written and oral exposition, see it through to completion, stand behind it, and put something new—and crafted by themselves, in community—into the world.

Off-loading the creative work, the heavy lifting, to a machine can be debilitating for the student.

That is a creative thing. The paper they have produced is a creation, but in creating it, they are also creating themselves—as historians in their own right, and as people who can grasp what people in the past went through and express why it matters. That is not something machines can do for them, though machines can and do help. But off-loading the creative work, the heavy lifting, to a machine can be debilitating for the student, rather than body building; enfeebling, rather than edifying and contributing to growth; atrophying, rather than developing the ambulatory practices and skills, intellectual muscles and sinews, needed for the journey.

Even without the looming presence and insistent pushing of generative AI, ask me at certain points in the process about getting rid of the research paper, and I might say, “Let’s throw it overboard and do something different, more accessible, and more like where history in the public is going.” But in retrospect—a perspective that comes naturally to us as historians, as we look back in order to chart the future—I still think that requiring our majors to write the research paper is a good foundation for their futures, and by extension our futures.

As I put together my pedagogical road map for the territory ahead, in which generative AI will be a ubiquitous feature of the landscape, I rely on history. In particular, the history of technology. That field is part of my origin story as a historian, for I wasn’t a history major as an undergrad. But a history of technology course ended up paving the way for my future. I was thinking about that course and its readings last year when a student chose the building of the Satsop nuclear power plant in Washington as his thesis topic. The cooling towers still stand as relics, but the plant was never completed. They were part of a program called the Washington Public Power Supply System—its acronym, fittingly, is pronounced “whoops.” The student, having driven 300 miles from Tacoma to Pullman to examine archival sources, wrote an illuminating thesis focusing on how labor, politics, and the environment all factored in to seal the nuclear plant’s unfinished fate.

His project got me thinking back on a then new book we read in that old course I took as an undergrad: Langdon Winner’s The Whale and the Reactor (1986). Winner raised deep political and environmental questions about nuclear power plants and other technologies. Decades before the appearance of ChatGPT, Winner presented a skeptical view of the hype around so-called artificial intelligence and what it promised. Hype, he claimed, provided “much of the persuasive power of those who prematurely claim great advances in ‘artificial intelligence’ based on narrow but impressive demonstrations of computer performance.” And here’s the kicker: “Children have always fantasized that their dolls were alive and talking.”

But in our era of ever more powerful LLMs, that fantasy is now easier to maintain—a mass hallucination, you might call it. In Autonomous Technology (1977), Winner pointed out that the history of technology is often the history of people releasing “powerful changes into the world with cavalier disregard for consequences; that they begin to ‘use’ apparatus, technique, and organization with no attention to the ways in which these ‘tools’ unexpectedly rearrange their lives; . . . that they endlessly proliferate technological forms of life that isolate people from each other and cripples rather than enrich the human potential.”

Based on this view of the technological past, and the legacies that have been carved out for us, I’m trying to apply a precautionary principle with respect to the introduction of generative AI to our learning environments. I worry that the drive to put machine learning into everything we do in education will one day appear, in retrospect, as analogous to the decision to put lead into gasoline in the 1920s. Sure, that additive addressed the knock in automobile engines, allowing us to speed faster down the highway encapsulated in quietude. There was a safe alternative: Ethanol additives have the same effect. But ethanol could not be patented; tetraethyl lead could be, and so the profits could be privatized, and the costs externalized, out of the exhaust pipe. This nation got in its cars and drove everywhere—and polluted the environment, spewing lead out into the places where we live. While severe lead poisoning can result in coma or death, chronic exposure can cause headaches, constipation, and the loss of memory and intelligence. We moved faster but thought slower. We blew past ourselves.

We moved faster but thought slower. We blew past ourselves.

The most extreme version of this lesson from history can be found in Caroline Fraser’s book Murderland, which traces and indicts both leaded gasoline and our local Tacoma smelter’s persistent plume for spewing lead into our environment, creating conditions on the ground that aided and abetted the rise of serial killer mentalities among us. Expose children to lead, and they can become “irritable, nervous, inattentive, slow to learn.” They can have hallucinations. They lash out. In the worst cases, they can die. Or the lead poisoning can be one factor in spurring them toward enacting violence on those around them. Ted Bundy, who briefly attended my college in the 1970s, grew up immersed in this environment.

As we navigate the intrusion of generative AI into our writing and teaching, we can draw on the history of technology to help us see the risks associated with hyped progress. The suffusion of machine learning into our total lives is a massive experiment. We and our students are its subjects. There are indications that AI use quickly becomes a threat to our human, natural intelligence and its continued cultivation. Our history with the automobile, including leaded gasoline, tells us that we have in the past cruised by the wreckage caused by new technologies while reveling in our newfound speed. In our ongoing experience with generative AI and all that it promises for education, the costs may lie in the debilitation of our students’ capacities for critical thought and expression, and a general toxification of our learning environments.

By the sweat of their brows, seniors last semester generated a wonderful set of original papers adding to the world’s knowledge, on topics ranging from how the monarch butterfly took flight as a symbol of immigrant rights to gender-empowering performances in Trinidad’s Carnival. During their research journeys, they transformed themselves from students absorbing history to historians questioning it and forging new understandings of the past. Along the way, one student took a detour through unauthorized use of an LLM to generate parts of an early draft. This became a learning opportunity, as we worked together in the middle of the term so they could ultimately create something new and of their own from the research they had assiduously done. Another student, inspired in part by their own experience with interest in an automobile purchase, examined the history of buying cars on credit. We are now in the midst of a new phase of turbocharged consumer culture, but we have the power to decide how much credit to extend to vehicles of all kinds running on AI.

We don’t know where this is heading. In another book we read in that history of technology class long ago, the odd third-person autobiography called The Education of Henry Adams, Adams reflects on a world’s fair he had recently visited. It was one of those “magic lands”—as John Findlay, one of our history majors who went on to become a renowned historian of the American West, calls them—where stories are told about ourselves, and where we are going, and that almost always put on display technologically propelled utopias of the future. As Adams saw it, “Chicago asked in 1893 for the first time the question whether the American people knew where they were driving.” That same question is before us now.