Thinking only through Models

“Critical thinking” is an extremely popular phrase, not only in the academy, where it has been institutionalized as a kind of Holy Grail that we are to lead students to, but in the world of online political battles, where it is invoked most often as that which one’s opponent is lacking (second only to “knowledge of history”). I’ve always resisted the term, first of all because, when pressed for a definition, its proponents invariably provide a list of banalities that would themselves need to be defined in terms that would only incite more definitional conflicts (“reaching conclusions based on evidence”; “taking into account other viewpoints,” etc.) but, even more importantly, because I’ve never seen anyone provide a clear genealogy of the concept, or have been able to get very far in doing so on my own, suggesting to me that “critical thinking” is a purely bureaucratic invention, much like “diversity” (there is no great thinker of either of these concepts). John Dewey seems to have used the phrase (which is not a recommendation for me), and there’s the concept of “critique,” with a clear point of origin in Kant, but discussions of “critical thinking” never refer to Kant, Kantianism, or the post-Kant trajectory of “critique” (in Marx, etc.). “Critical thinking” in every single one of its uses really means something like “thinking like me,” and it would be more honest and illuminating to make that explicit, since presenting ourselves and treating each other not just as people with opinions and ideas but as models of thinking (that we at least implicitly want to propagate) would bring our encounters into much clearer focus and provide for much more verifiable observations regarding the thinking process than going through the checklist provided by “critical thinking.”

Part of the effect of centering “critical thinking” is to detach us from texts and thinkers that compel us so as to subject us to bureaucratic testing methods and ideology. Students obviously learn much more from entering, even or especially naively, into the “logic” of a compelling text and taking it on as an identity, a position from which you see everything, going beyond the model itself, so that you’re trying to figure out and are ready to insist upon what, say Nietzsche (who is I think a very typical thinker in terms of inducing such an experience) would say about bowling. Entering and “uncritically” adopting a text or author in that way will ultimately lead you to its limits (in which situations does Nietzsche not, in fact, provide an answer?), and will also lead you to other thinkers whom you could merge with in turn, carving a distinctive path through intellectual history. When you get to the point where you feel confident about asserting how Nietzsche, Tolstoy, Kant, Melville, Heidegger, Gertrude Stein, etc., would take up a particular question differently, you will find attempts to apply the institutional criteria for “critical thinking” as annoying as I do. Your thinking will be “partial” and idiosyncratic, but no less than anyone else’s, and you will always be in good company.

This argument leads to the Girardian question of which models to choose, but that’s a problem as well since first of all our models choose us and then guide us towards subsequent models—that is, from what model-free position is one “choosing” models? The pragmatic resolution of this paradox is a pedagogical one—at a certain point you will intersect with others carving their own path through intellectual history and be confronted with the reciprocal imposition of models that will shape that encounter and it is then that you will propose models for the other and in this way find yourself backtracking through your own intellectual history and providing various weights to your models. We then have something like a disciplinary space and how weights are assigned to various models, or differing versions of those models, will be determined by maintaining the disciplinary space, or the conversation. Once you are thinking about producing models for others the question of which model to choose is no longer situated within individual contemplation but in the continual testing of “merges” (I’m using the word “merge” with an inflection taken from Eric Jacobus’s ROBA model) and while this is no guarantee of “correctness” it does provide for a kind of responsibility that gets you into the habit of thinking through the implications of modeling yourself one way or another. Walter Benjamin was right to say that we are incapable of resenting those who will succeed us (because, in my view, we can’t flesh out the world in which we’d have to imagine ourselves competing with them) while we are certainly not only capable of but almost addicted to resenting the past, so this shift from thinking about which model to adopt to thinking about which ones to produce represents entry into a higher ethical world.

I’m drawing here on many previous discussions (with “Exhaustive Imitation” probably the most recent) calling for the maximization of imitation, first of all by tearing down all the pretenses leading us to imagine we could ever be doing anything else. If you see yourself as taking a “critical distance” from all models and assessing them all according to some epistemic or ethical criterion that just means you’re modeling yourself on some institutional figure who represents the power behind that kind of assessment. My argument for exhaustive imitation follows from my longstanding adoption of mimetic theory but it is also a way of backtracking my adoption of mimetic theory by identifying the intellectual habits that “primed” me for it. This, indeed, was my only way of learning for a very long time—enter a compelling model thoroughly, live it, apply it to everything, stand ready to defend it against all opposition, trace it through its predecessors and successors as best I could until the next compelling model came along—and what counted as a compelling model was whatever most effectively shattered my “internalization” of previous models. Then you get to the point where you can do this in a more controlled way, especially once you understand what, exactly, you are doing and bracket the shame involved in understanding that. If I had sufficient means, I could work through each sentence I’ve written here and provide a genealogy for my acquisition and use of that word through that ongoing but now systematic and controlled battle of the models through which I learned to see through a thousand eyes, most of which I could not remember or identify. (There’s a kind of reading of Husserl—whom I have in fact only “read” through others—in the phrase “bracketing of shame,” which for me situates his phenomenological concept within an anthropological/mimetic frame.)

Exhaustive imitation is worth returning to (there’s a point at which you can take your own prior iterations as models) because it places us right in the middle of problems posed by AI or the Large Language Models which are “imitation games.” To get a machine to imitate you you have to look at what you’re doing in a fresh way, breaking down your motions intricately because operationalizing them computationally will always have to be done indirectly through a process of breakdown and reconstruction. The obsession with distinguishing humans from machines leads to a defensiveness that militates against taking this new opportunity to re-examine the human, leading to the paradoxical result that lists of things humans can do that machines can’t merely provide a menu of AI improvements for the next generation of programmers to attempt. I didn’t develop the notion of “originary debt” with AI in mind but I think it can very effectively be pressed into service here insofar as it, like any other distinctly “human” attribute, presents a challenge for programmers to install it algorithmically, but also ensures that such programming efforts will run parallel with rather than replacing our own working out of our own debts—the reason for this is that the technological developments themselves continuously deepen and modify the debt by transforming our relation to our successors, to whom we try to transmit the debt so as to approximate the convergence of debt enforcement and debt forgiveness. (Think in terms of making provision such that our descendants’ absolute fidelity to us will simultaneously entail a plasticizing of the models we transmit.)

I’ve come to see the “vocation” or best use of the LLMs to be the prolongation of intention. I wrote at some length about the question of “intention” in an essay I recently published in an academic journal, but to put it briefly (and then get to a conclusion I hadn’t quite yet arrived at then) here: “intention” is joint attention directed towards the maintenance of linguistic presence, which is always at risk or, even better, always already under siege. So, when we’re speaking about the intention of a writer what that means is not something like “what Shakespeare is really trying to say here is…” but something more like “Shakespeare is engaging and exploiting the conditions offered by the Elizabethan theatrical world (set within the Elizabethan world) so as to ‘re-scene’ those scenes in as iterable a way as possible.” To take a more commonplace example, the intention of a student writing a paper is to position the reader of that paper (the instructor) as favorably toward the student as possible—even if the student couldn’t really say exactly that to instructor or maybe even to himself. This recognition of the undeniable reality frees us from pretending that the student is thinking through and wishes to “communicate” “his” ideas to an imaginary audience and enables us to engage more explicitly our real institutional situation—the instructor can proceed on the assumption that the student wants to know what the teacher wants and use that assumption to get more explicit and “intentional” in communicating what he wants. We can then perhaps come closer to sharing the intention of making the historical and institutional space of the classroom as productive as possible, making that setting, within the disciplinary assumptions of the academy and the social function of the academy, the “topics” of the writing.

The concept of intention, which can always be reduced to the simple “what do you want?” while being expandible into the infinite conditions that would be drawn into the articulation of that want, therefore enables us to connect the basic mimetic and “populist” question, WWXD? with a persistent questioning of the distance between us and whoever that X might be. In other words, the prolongation of some intention that we want to bring to bear on our work is what informs the prompt. What, for example, would George Washington make of our sprawling administrative state, accountable to no one and field of competing lobbyists pressing bureaucratic buttons to receive some return for their clients? This becomes a very challenging research question, because we’d have to figure out who, exactly, was George Washington (and which texts, of Washington and others, do we draw upon, and how do we weigh them, in determining this), or, perhaps, which George Washington do we wish to prolong into the present; and, then, how does whatever we are to take Washington to have been precisely in his difference from any of us or anyone we could imagine today, get mediated in his Rip Van Winkle awakening into a no doubt confusing and in many ways appalling (but also amazing) world—do we want to account for that disorientation as well, so as to draw out that in his intention towards, say, the office of the presidency, that we wish to converse with and model ourselves on? (The challenging research questions overlap with formulaic sci-fi time travel movies, which we would want to control for without denying.)

Let’s break down imitation into an attentional model, with single attention transformed into joint so as to complete the imitation. If you pay close enough attention to someone, so as to virtually hypnotically be absorbed in their movements such close following eventually requires you to move your own body as closely in sync with the other simply in order to continue tracking them with the closeness your absorption demands. Imitation, then, grows out of attention and then joint attention emerges from the stand-off that results from imitation so close that the two converge on the same object—imitation can then only continue without canceling itself in a struggle that would put the two decidedly out of sync with each other if that convergence swerves into the gesture of aborted appropriation. It is this gesture that we are prolonging in the maintenance of linguistic presence, while any particular figure, like George Washington, is a way-station, but an absolutely necessary one because the prolongation of this gesture is human history. In any encounter we always have in mind various possible disruptions of presences that we might need to fend off, with “back-up” modes of presence in case the fending off fails, and so on, and it is here that we can use a curated body of texts against which we have trained our model to generate unlimited scenarios that serve pedagogical purposes, i.e., the construction of artificial scenes in which various paths from attention to imitation to joint attention can be simulated and the results recorded and fed back into the program. This approach would fulfill Nietzsche’s view of the “use of history for life” and make real all those familiar and, of course, “cheesy” calls to imitate this or that admirable figure—those calls to imitate admirable figures are completely correct, with the “cheesiness” merely a result of the inadequate and narrowly subjective means of determining WWXD we have so far had available. We’d be using the past to fling ourselves further into the future while maintaining massive research projects that are focused on the most important thing—the retrieval and reconstruction of the intentions of all of our ancestors from whom we might like to hear from—it would even be a recovery of ancestor worship, no doubt the most ancient form of ritual, the traces of which remain evident everywhere.