Critical Writing Literacy

I have been thinking a lot about critical writing literacy, and I’m trying to get it into a paper. We’ll see how that goes. The concept comes out of critical literacy, which you might have heard of. Critical literacy is concerned with the fact that texts do things in the world. They come from somewhere. You know, they have readers in mind, even when the writer has not thought carefully about who those readers are, and they carry assumptions. When we bring large language models into this conversation, we usually turn our attention toward the text the model produces. We worry about accuracy, bias, hallucinations, source material, all of that, and rightly so. I am interested in turning the lens around a little bit. Once I start using one of these tools in my own writing, what exactly am I doing? What have I decided the tool can do for me? More importantly, because this is where I think the question becomes less abstract, what have I decided I do not particularly need to do myself?

Writing is not just, well, “writing,” as you will know. We use the word as though someone sits down and performs a single activity called writing until a document appears in a puff of smoke, when in reality there are all sorts of things happening under that heading. I am trying to figure out what I think. I am trying to get some version of it onto the page before I forget where I was going. I come back later and discover that what seemed perfectly obvious yesterday is barely intelligible. Then I start worrying about a reader, or I realize that I have written three paragraphs before getting to the thing I actually wanted to say. Somewhere in there I start fixing sentences, which I probably should not be doing yet, and then I notice a citation and disappear into Zotero for ten minutes… you know how this goes.

Even the familiar language of audience and purpose already tells us that writing involves judgment all the way down. There is no neutral act of putting words on a page. So once AI enters the process, I do not think the interesting question is simply whether I used it. I want to know where it entered, what problem I thought I was solving, and why that particular bit of the process struck me as something I was willing to give away.

Okay, smart guy, so you’re saying we should be more mindful when we are writing. Duh. I mean something a little more concrete than that though, and one need not smuggle in anything overtly political to get there. Take the old advice that drafting and revision should be separated. You have heard versions of this for years and still routinely ignore it, don’t you? You write a sentence, reread it immediately, fiddle with a word, decide the sentence is ugly, start worrying about the paragraph, and twenty minutes later you have polished six lines of prose attached to an argument or goal you have not actually worked out.

Drafting asks for a certain willingness to tolerate provisional language because you are still finding out what you have to say. Revision asks you to return with enough distance to notice that the thing you produced does not quite do what you thought it did. Those are different kinds of attention. The distinction is important with AI because a model can make provisional language look right very quickly. You can have a paragraph that sounds correct before you have done all the thinking you would normally do.

This brings me to the part I care about more, which is what we are practising while all of this is happening. Every time I work through a difficult piece of writing, I am getting slightly better at some part of the work (I hope!), assuming I am paying enough attention to learn anything from it. If I keep trying to explain an idea to a reader who does not share my background, eventually I get better at noticing where the explanation is a bit wonky. If I have to organize a messy argument myself, some of that organizational judgment carries forward. I do not mean this in a noble “the struggle makes us stronger” sense. Plenty of struggle is stupid. I have spent embarrassing amounts of time formatting references, and I would happily allow software to eliminate that work. The more interesting issue is that we have limited time and limited attention, so we are always deciding, usually without saying so, which capacities deserve more of us. I use a calculator and feel no anxiety about what this has done to my long division (well, mostly). That was a choice, although such a normal one that it barely registers as a choice anymore. AI makes much more consequential versions of that decision available.

And then there is a selfish dimension to the whole thing that I think gets obscured when the conversation stays at the level of learning outcomes. Suppose I keep handing a particular kind of writing over to AI. The standard response is that I may fail to develop the relevant skill, which is true enough. I am increasingly interested in the prior question: do I even want that skill? Take grant writing. There are good reasons for researchers to learn how to explain why their work matters to people outside a narrow specialty, and grant writing can force you to do that.

At the same time, someone may discover that they care intensely about the research and feel much less enthusiasm for the performance surrounding the competition for funding. They may dislike what happens to their thinking when every idea has to become a significance claim or a promise about impact. They may simply be tired of the whole business. If that person starts giving more of the grant to AI, I would not immediately treat this as evidence that the tool has corrupted their scholarly development. I would want to know why the delegation came so easily. Maybe they are avoiding a skill they need and do not yet possess. Maybe they have begun to disengage from a part of academic life they once assumed they were supposed to value. Those are different situations, and the second one could tell them something fairly important about where they are headed.

I have a much smaller version of this with briefing notes. I do not especially like writing them. I understand the genre and I can see why someone in policy would care a great deal about becoming good at it. In that setting the genre is doing something valuable, perhaps, and I imagine I would feel differently about learning its conventions if policy work were central to my professional life. In my current role, though, briefing notes can feel strangely disconnected from what I think I am there to do. They can start to feel like institutional work produced because institutional work tends to produce more institutional work. If I reach for AI almost immediately, I could lecture myself about maintaining my ability to write briefing notes, and maybe I should. I could also take my own lack of interest seriously for a moment. Why do I care so little about getting better at this? Is the task genuinely peripheral to the work I value? Am I being precious about my role? Am I calling something busywork because I resent being asked to do it? I do not necessarily need to resolve all of that before I open ChatGPT. I do think the speed with which I want to offload the task is information about me, and I lose something if the only question I ask is whether the resulting note is good enough.

A journal article makes the same issue messier because other people are involved and because my relationship to the work can change halfway through. Imagine a colleague asks me to help with an article under a tight deadline. I agree because I like the colleague and want to help. The research area no longer means very much to me, which I know going in, although I find that the drafting is surprisingly enjoyable. I remember how to think in that literature. There are old muscles there. Then the deadline gets close, my kid gets sick, the rest of life continues to happen, and suddenly the revision has become a much larger claim on my evening than I had expected. So I use AI. I am careful about it. I give examples of my writing. I explain what kinds of changes I want. I try to stop the model from sanding away too much of the prose. In other words, I do all the responsible prompting people recommend, and the result is technically fine. Cleaner in places. More consistent. Easier to finish on time. Then I read the whole thing and it feels cold and dead. I recognize the argument. I can defend every claim. I also feel as though somebody has removed me from the sentences while leaving my name attached to them. A weird kind of psycho-social decapitation.

I could still submit that article, and I might. My colleague gets what they need, the article gets finished, and my kid was sick. There is no reason to turn the decision into a little morality play about authentic authorship. What catches my attention is my own reaction to the result. Apparently I care about sounding like myself even in an article about a subject I no longer care much about. Or I care up to a point, because under enough pressure I was willing to accept prose that I experienced as less mine. That sounds inconsistent, which is probably why I find it interesting. I valued helping the colleague. I valued the time with my family more. I valued my voice enough to notice its absence and not enough, on this occasion, to spend another four hours restoring it. The AI did not create any of those priorities. Its presence gave me a very easy way to act on them, and then the resulting text showed me the trade I had made.

I think critical writing literacy has room for this kind of selfishness. We are often quite good at talking about what students should learn, what scholars owe readers, what counts as responsible use, and I understand why those questions dominate institutional conversations. A person also has to decide what kind of work deserves their finite attention. You cannot maintain every skill at the same level forever. You will get better at some things because you keep doing them, and other abilities will become rusty because you stopped caring enough to practise (like my dabbling with poetry). Some of those choices will be mistakes. Some may come from exhaustion or resentment and deserve another look later. Others may represent a fairly accurate recognition that your work has changed and that you no longer want to become the person a particular genre is trying to train you to be. There is something strange about assuming that every skill offered by professional life deserves preservation simply because it is a skill.

AI makes these choices easier to act on, which is part of what makes the technology compelling to me. I can now avoid parts of writing that I once had to do myself simply because there was no practical alternative (e.g., coming up with titles). Sometimes I am delighted by that. At other times I get a document back and notice that I have removed the part I cared about, or that I have used efficiency to wallpaper over a problem I would have been better off confronting and sitting with for a while (slow academia anyone?). And sometimes the document is fine and I am fine with it and there really is nothing more profound going on. Critical writing literacy, at least as I am thinking about it, leaves room for all of those cases. I want to become better at noticing what I am handing over and what made me want to hand it over, especially when the answer has less to do with AI than with my changing relationship to the work.