Show full content
Life is change
How it differs from the rocks
I’ve seen their ways too often for my liking
New worlds to gain
My life is too survive
And be alive for you
From Crown of Creation, Jefferson Airplane
A couple of weeks ago, we had the release of “Einstein Companion,” which the Ars Technica coverage trumpets “Doesn’t Just Help With Homework. It Takes Over Your Role as a Student.” It apparently connects to Canvas, the learning management system (LMS) that my university uses, and in which I currently have about 290 students for Anthro 101. If the corporate propaganda can be trusted, it can read documents, including the course syllabus, watch videos, interpret slides, and complete and submit assignments, on time, on behalf of a student.
Meanwhile, unbeknownst to us professors until after the fact, the university rolled out enhanced AI tools to students toward the end of January (as far as I can determine.) When they open a Google doc now (which I use for assignments in my classes), they are greeted by a very large, prominent “ask Gemini” button at the top, a “help me write” prompt right at the cursor, and another in a context-aware menu to the right. It offers to help them “refine” a passage, “expand” on it, make it “more formal” (or less), etc. Those buttons and menus are cluttering up my screen even as I draft this post (sadly unaided).
At about the same time, OpenAI CEO Sam Altman was asked about the energy cost of training and running Large Language Models (LLMs) like ChatGPT. He responded, more than a bit defensively, that humans too, are expensive to train and run. For example, you have to feed them for twenty years “before they get smart.”
We happened to be discussing human exceptionalism in class that week, and that Altman clip generated a lot of discussion. If we’re down to discussing the energy value of humans relative to LLMs, “what are we even doing,” asked one student.
Then a colleague, Alberto Acerbi, a leading scholar in the modeling of cultural evolutionary processes, posted a thread on X (erstwhile Twitter), in which he says that he asked Claude Code (another LLM) to elaborate on one of his own pre-prints. Claude wrote a paper, complete with simulation experiments, analysis etc, which Acerbi reports is publishable with minor tweaks. “…it would be possible,” he says, “for someone to create hundreds (?) of those with small efforts.“ He ends the thread by asking, much like my student in 101, “what should we do?”
To top it off, the CEO of Claude’s Maker (I suppose I should capitalize that), Anthropopic, warns of “an impending AI tsunami that will upend human society as the tech surpasses human intelligence.”
What should we do indeed? Even absent the predicted rogue wave of computer-based super intelligence, in this new world of good-enough automated writing and instantaneous analysis, wither academia? I’ve been asking myself that question for a while now.
LLMs can already write papers better than most humans, certainly better than most undergrads. They can draw better than the vast majority of humans. They can search parameter spaces faster and more completely than any human researcher. They can write and review articles, code simulations, analyse results. Using NotebookLM, they can even learn the content of textbooks and courses, and produce interactive podcasts. They can give our courses, live and in person, any time, anywhere, at much lower cost than we can. Our final preserve it seems, the Human of the Gaps, is that we can take responsibility for their work and their decisions, something they legally can’t do themselves at the moment. But how many of us are needed for that? And who would be willing to do it?
Human learning
In the face of all this, the only thing we can do is remember who we are, and what our mission is. We are an academic community. Our mission is to learn about the world and to help others learn.
We can also benefit from remembering what we’re not, or at least what we shouldn’t be: We are not an article production machine. We are not a student ranking engine, either, although that’s what many would like us to be.
In class, I remind students that we are here to learn, together. There is no future for them without AI (except perhaps a very depressing, post-apocalyptic future, which this morning looks increasingly likely), and they will have to learn when and how to use it, but especially when and how to not use it.
I don’t tell them not to use AI. I don’t make AI use into a moral problem. It is a practical issue in learning. I simply tell students that if an LLM does their course work for them, they are not learning. They are only hurting themselves, not to mention wasting time and money. There are plenty of good, productive, ways for them to use AI. Doing their course work for them is not one of those.
I tell them that my goal is to help them learn, and to learn with them. My goal is to help them prepare to go out into the world and make a contribution to their community, however they define and identify that community. Over the years, I’ve learned a few things about contributing to a community, and I want to share that with them. That message seems to resonate.
In terms of assessing student learning, which is different from ranking, we can either get into a losing arms race with software like Einstein Companion, playing increasingly futile digital whack-a-mole, or we can change our approach to emphasize more human assessment of human learning, in a world increasingly dominated by computer assessment of computer work.
We can make sure students at all levels understand that they are part of this academic community of ours. They are welcome here. They can contribute. At the end of the day, students will do their own work and their own learning if they want to, if they think it is worth doing, and if they value both the process and the outcome.
We need to engage with our students and their learning as directly as possible, as fellow learners. We are learning with them. If their goal is to get a credential so they can take their place as a cog in a faceless socio-economic machinery, “in the world of bigger motorcars,” as Jethro Tull’s Thick as a Brick once put it, which is how many of them view things at the moment, they will use whatever machines are available to them for that purpose. I can’t blame them. Born forty years later, I would probably be doing the same.
If, on the other hand, their goal is to learn, they will learn. We create the conditions under which they labour, and we help drive their decisions.
For a couple of years now, I’ve assigned a reading journal as the only work in my courses, which I think I will rename “learning journal” for future semesters. I have about 290 students right now. I comment on each and every reading journal. My plan this semester is to comment on them three times each. I meet with the students. Not much. Just little fifteen minute meetings. It turns out that it’s enough for me to engage with them and with their learning. Some students want to meet. Some don’t. That’s fine too. Different engagement strategies work for different people. I make the time to meet with those who want to.
The outcomes are encouraging. I see very little use of AI, because students know that I am interested in what they have to say, and I am interested in helping them answer the questions they have about the world, through an anthropological lens. I treat them like human beings and like fellow learners, not like LLMs and not like exam completion algorithms. They reciprocate.
In Anthro 101, I am not so interested in whether they know the conventional (and ever changing) dates for the initial Out of Africa dispersal, for example. I tell them that I am interested in seeing how they can develop the beginnings of an anthropological perspective on the world.
I’m not interested in the finished product, in the polished answer, which an LLM can create much faster and better than most of them. I’m interested in the process. I am interested in the getting there, more than in the knowledge itself. (And here, I find myself consciously trying to avoid the now classic LLM tell-tale formula of “it’s not x, it’s y.” Already.)
And what, they ask, is that anthropological perspective I want them to develop? Here is what I tell them: Always consider the whole human, biological, cultural, and linguistic, past and present. Develop a basic awareness that human variability in all those dimensions is greater than they have ever imagined. Acquire a sense of cultural relativism that acknowledges its own limits, a consciousness that humans are part of a larger web of things, organic and inorganic, living and not. There, congratulations. You have now taken my Anthro 101. Now you just have to do the reading, and tell me about it from those perspectives.
All this, including the reading of the journals and the individual meetings, actually fits very well within the forty percent of my job that I am supposed to be spending on teaching. Forty percent is a lot of hours in a semester, especially when I spend most of my time on other things outside the regular teaching term.
I even did the math myself, since LLMs are not very good at that part of it yet. If I have 250 students (which is what I recommend), and if I read their journals three times (15 minutes per read on average) and meet with each for 15 minutes, that’s approximately 20 hours per week during the semester. Add six hours for lecture and course prep, and pro-rated over a year, that’s less than the famous forty percent of my typical work week. I even have time to write the occasional blog post.
Not everyone, at every stage of life and career, should do this. Communities are diverse, and there are diverse ways of contributing. This is also part of being human. But this is what I can do now. I could do even more if we decided that we are not prisoners of the course and semester structure that we’ve imposed upon ourselves, almost completely because of historical contingency, and for reasons of administrative convenience.
Yes, it’s a lot of work. But I can’t complain about the pay, and most of it comes from the public, so I feel that I must give their children their money’s worth. It’s also endlessly fascinating and stimulating work. I see a student’s perspective shift within a semester as they read the course materials (and the evidence says they do), as they make connections with material from other courses (learning across the curriculum), with events in their lives, and even in the world around them. Best of all, I have the evidence that all of this is happening, in their own words.
With their help, I glimpse regions of the discipline to which I have had little, if any exposure. I help them explore as much as I can, by giving them comments, starting points for further reading and reflection, tips on writing (so many tips on writing, which is how I know they are not using AI).
Even those international students who tell me they initially use AI to clean up their text eventually stop, mostly. I make sure to tell them that I am interested in their thought, their writing, their expression. I am not going to punish them because English is their second, or third, or fourth language. I judge not, lest I be judged, especially as a second language writer myself, I tell them.
On the contrary. I am interested in how they express themselves, in their linguistic peculiarities and individualities, because I am an anthropologist, and I consider the world from that anthropological perspective that I want to share with them, and that I want them to start developing and using.
And if a few of them fool me, if they train an LLM to write like a first year student whose third language is English and who went to school in Chengdu, or like a fourth year psychology major educated in rural Alberta (of which I have many this semester), in the end, the loss is theirs. The vast majority will have learned more about being human, and they will be better prepared to meaningfully participate in a community. Good for them.
Human scholarship
These principles all transfer quite directly to the scholarship part of our workload. Many tools are available to us for pursuing that mission. The LLM is now one of them, and soon, no doubt, other forms approaching Artificial General Intelligence (AGI). Archaeologists have always been notorious early adopters. From the theodolite used by William Stukeley at Stonehenge in the 18th century (page 51), to aerial photography as early as the First World War, to genetic algorithms in the 1990s.
Like all of these, LLMs have their uses, and they will be used in scholarship. AI is great at finding patterns in landscapes and assemblages, for example. And if LLMs can draft some boiler plate for us to edit, so we can spend more time thinking about our data, so much the better. Then of course, we shouldn’t fault our students for wanting to use the same strategies.
Just as our students respond to the learning environment we create for them, the debilitating deluge of AI paper submissions to journals will only materialize if we don’t make necessary changes to our academic culture.
Quantity has never been a good metric for scholarship, and as Acerbi points out above, it is now completely obsolete, and even actively harmful. So what do we evaluate then, if not numbers of papers published and numbers of citations?
Perhaps we do something radical and evaluate contribution. Does this contribution help us learn? Does it help us figure things out, and answer questions that we care about answering, as a community of human beings? This means we need to read and evaluate.
If we’re going to actually read the work of applicants to grants, jobs, graduate programs, promotions, pay increases etc, which is mostly when we rank, we will have to ask each other to produce and present less, not more. But what we produce should matter, and it should be recognized as such.
Yes, LLMs can produce millions of papers, as quickly as we give it chips, water, and electricity for. A percentage of those papers could even represent potentially useful contributions. But they will only be useful contributions to us as humans if we can read them as humans.
Without forgoing the tools that make our work faster and more efficient, that allow us to search through databases at speeds that would leave the heads of our grandparents spinning, that crunch numbers and yes, even string words words at undreamt of rates, we will have to produce work that can matter at a human scale, because we are a community of humans.
We will have to embrace our limitations relative to LLMs, and be aware of our unique features. We will have to finally develop a scholarly landscape in which quantity is irrelevant above a certain reasonable threshold, and in which contribution is measured by real world impact on real world communities.
The question should never be how many papers, or how many citations, and certainly not in what journal (which heaven save us from). We won’t benefit from counting each other’s productions. We may, however, benefit from learning from each other’s contributions to knowledge and practice.
The question should be what difference does, or could, this work make for us, as a community of humans living in a larger world. It’s more work than counting outputs. It’s more reflection. It’s a greater ethical struggle. Like reading student learning journals, it means more engagement. But like reading student learning journals instead of applying an answer key to a quiz, it’s worth it in the end.
At the end of the day, we’re all humans, learning to be human in a larger world. We are members of communities, and we should want to contribute to our communities, to make them better for all of us. We do this by learning and by doing. Only we can do our learning. Various AI tools can help us with the doing, no doubt, but we still have to do the learning, together. As Jefferson Airplane told us more than fifty years ago, before the LLMs arrived, our life is to be there for each other.










