Tertiary algorithmicity, the bypass of cognition, and the case for pedagogical friction in an age of generative AI.
This conceptual paper argues that generative artificial intelligence requires an extension of Walter Ong's media ecology beyond primary orality, literacy, and secondary orality. Drawing on Ong's account of how communication technologies restructure consciousness itself, the analysis identifies three assumptions in his framework that no longer describe the current media environment: that humans create symbolic content, that distribution follows relatively transparent logics, and that consciousness encounters media as an external environment.
In their place, the paper proposes two stages. Algorithmic secondary orality names the transitional phase in which platforms organized — but did not yet author — symbolic content. Tertiary algorithmicity names the present condition, in which algorithmic systems both create and curate symbolic content, making human authorship optional at scale. The condition produces three displacements that matter for learning: noetic displacement, rhetorical saturation, and existential abstraction.
Read against decades of learning-science research on productive struggle, retrieval, and the generation effect, tertiary algorithmicity threatens to normalize what Kapur (2016) calls unproductive success — fluent performance without the cognitive labor that produces durable understanding. The paper develops pedagogical friction as a response framework, with four dimensions — noetic (the head), rhetorical (the room), existential (the world), and infrastructural (the system) — and distinguishes between friction that builds capacity and friction that excludes.
Ong's account traced three stages in how dominant media restructure the noetic world. Two additional stages are now required to describe what algorithmic curation and generative AI have done to the symbolic environment.
The case for extending Ong rests on a hinge. If these assumptions hold, generative AI is only an intensification of secondary orality. If they break, new conceptual categories are required.
Outputs are not transcriptions of human thought but statistical predictions of plausible text. The source of symbolic expression no longer must be human consciousness.
Two users opening the same app at the same moment encounter different content, sequenced by computational predictions about engagement. The shared symbolic environment fragments.
The symbolic world a learner inhabits is increasingly shaped by systems that infer attention, predict preference, and return those patterns as a self-reinforcing loop. The book on the shelf has become a mirror.
Each names a different way the learner is repositioned under generative AI — as thinker, as interlocutor, and as accountable claimant. A student who submits AI-generated work experiences all three at once.
Noetic displacement names the condition in which the externalization of cognitive labor moves from storage to generation. Operations that previously had to occur within the learner's own noetic activity are increasingly offloaded to an external socio-technical system.
Literacy externalized memory to the page. Secondary orality externalized broadcast. Algorithmic secondary orality externalized curation. Tertiary algorithmicity crosses a new threshold: generative systems produce new combinations of language that take the form of knowledge without the experiential grounding that has historically accompanied symbolic production.
Rhetorical saturation is the condition in which algorithmic systems produce discourse at scale, flooding communicative environments with synthetic text, image, video, and audio that is increasingly indistinguishable from human production.
Three features sustain it: origin uncertainty (humans struggle to distinguish AI-generated text from human writing at rates close to chance); simulation of expression (generative systems mimic distinctive style, voice, and apparent personality); and simulated dialogue (conversational interfaces produce the form of exchange without an interlocutor with stakes).
Existential abstraction names the condition in which symbolic production is severed from lived experience, situated perspective, and personal accountability. Text generated by a neural network expresses positions without having stakes, simulates conviction without being committed.
Ong traced a progression from orality's inseparability of speaker and speech, through literacy's authorial distance, to secondary orality's partial restoration of dialogic qualities. Tertiary algorithmicity extends this trajectory not to distance but to elimination. No consciousness stands behind the output. No person bears the intellectual risk of having committed to the claims the text advances.
Constructivist theory and the learning sciences converge: durable learning requires enough cognitive struggle to reconstruct existing schemas. Tertiary algorithmicity inverts this default.
Dewey, Piaget, Vygotsky, Kapur, Bjork, Sweller, Roediger: across distinct traditions, a converging finding — durable learning happens when existing schemas prove inadequate and must be effortfully reconstructed.
A condition in which the learner produces the fluent, plausible, often correct academic performance — without engaging in the interpretive and compositional processes through which understanding is built.
The educational threat is not that students will cheat. It is that schools will normalize a relationship between performance and understanding in which the two have come apart.
Three dimensions of educational threat follow from the analysis: cognitive (the bypass of the labor through which schemas form); rhetorical (the erosion of dialogic conditions that make ideas testable); and existential (the abstraction of authorship from accountability). These threats do not operate uniformly. Schools remain structured human environments, with teachers, relationships, and routines that can preserve the hooks for learning that broader media environments often dissolve.
Friction is the effort through which cognitive development can occur. Where frictionless automation is the default, the design problem becomes the deliberate preservation of the right kinds of difficulty.
An argument for preserving friction can easily become an argument for preserving inequity. The framework requires a distinction — and a question educators ask of every instructional decision.
The cognitive effort that builds schema, develops capacity, and enables transfer. Wrestling with difficult texts, defending arguments against genuine critique, composing original work, revising based on feedback. These difficulties are desirable in Bjork's sense.
Retrieval before AI assistance. Oral defenses of written work. Process portfolios that document revision. Peer critique of original drafts. Collaborative inquiry on contested questions. Reflection on the path of one's own reasoning.
Arbitrary obstruction that prevents participation without building capacity. Language barriers imposed on content assessment. Inaccessible formats. Procedural requirements unrelated to learning. Difficulty that treats one cognitive profile as the unmarked norm.
Dolmage (2017) argues that what higher education has called "rigor" has often functioned to preserve access for students whose embodied relationship to reading, writing, and timed performance already matches institutional expectations. Any call to preserve difficulty warrants scrutiny.
A student uses generative AI translation to convert an essay written in their native language into English. The friction bypassed may be exclusionary rather than productive: the language barrier prevented them from demonstrating content knowledge or participating in academic discourse.
At the same time, the translation process may bypass the productive friction of composing in the target language. The same AI assistance can be access for one learner and bypass for another in the very same classroom.
The pedagogical question is not "does this difficulty build capacity?" — but "for whom does this difficulty build capacity, and for whom does it restrict access?"
If tertiary algorithmicity makes cognitive bypass the default, then curriculum must make thinking visible, advocacy must defend productive struggle as an educational value, and policy must create the institutional conditions in which teachers can preserve friction without acting alone.
Process-oriented pedagogy assesses the quality of thinking rather than the polish of products. Sequencing matters: tasks that require noetic struggle before AI assistance preserve the generative cognitive work that prepares students for future learning. Oral components, peer critique, and written reflection introduce forms of friction that resist algorithmic substitution.
Educational leaders are positioned to name friction as a protected value. The dominant discourse frames innovation as efficiency, personalization, and scalability — but efficiency is the visible edge of a deeper economic logic. The advocacy task is to articulate why the cognitive labor of learning is not an inefficiency to be optimized away.
Most district and university AI policies address integrity, privacy, bias, and risk. These are necessary. But they are silent on learning science. Friction-centered policy makes explicit what cognitive processes education aims to develop, why those processes warrant protection, and what institutional conditions support educators in preserving them.
The argument is contestable. The framework remains valuable not because it answers every objection but because it makes the right ones visible.
No. Pedagogical friction is not a call to ban tools or restore some imagined simplicity. It is a recognition that learning requires certain kinds of difficulty, that those difficulties are what the learning sciences describe as desirable, and that under conditions of tertiary algorithmicity those difficulties have to be deliberately preserved. The framework is forward-facing; it asks what schools must protect under new defaults, not which technologies must be excluded.
Ong himself resisted this reading. Each transition brought losses as well as gains. Literacy enabled analytical detachment but diminished communal and mnemonic capacities. This paper retains that frame: tertiary algorithmicity is not a successor that renders prior forms obsolete. The stages coexist and overlap. The argument is descriptive of dominant conditions, not prescriptive of inevitable progress.
The disability-studies critique (Dolmage, 2017) is real and is taken up directly in Section V·C. The framework does not resolve the tension. What it offers is an interpretive structure that requires educators to ask, for each instructional decision, whether the friction in question builds schema the student would otherwise not construct, or whether it imposes barriers that treat a particular cognitive profile as the unmarked norm. This question has no universal answer.
The paper acknowledges this directly: generative AI can reduce exclusionary friction. The paradox is that the same tool that provides access for one student can remove productive friction for another. The pedagogical question — "for whom does this difficulty build capacity, and for whom does it restrict access?" — is meant to make that paradox a routine part of instructional design, not a special case.
They make the analytic move toward extension, but they treat digital interactivity as a continuation of orality's logic. Generative AI represents a categorically different condition: not only the conditions of distribution shift, but the source of symbolic production. Tertiary orality still presumes human authorship. Tertiary algorithmicity does not.
McLuhan provides one set of resources, and the 2024 special issue of Explorations in Media Ecology uses them productively. But McLuhan's framework is most powerful at the level of what media do; Ong's developmental account asks what they restructure inside consciousness. For an argument grounded in the noetic conditions of learning, Ong is the more direct theoretical ancestor.
Yes. The paper is conceptual. It does not present empirical evidence for how tertiary algorithmicity affects student learning, how educators navigate friction preservation, or how policy shapes the conditions for friction. These are the questions the planned dissertation is designed to take up.
This qualifying paper builds the conceptual apparatus. The dissertation puts it to empirical use across institutional contexts.
The framework — extended Ong stages, four dimensions of pedagogical friction, the productive/exclusionary distinction — generates specific questions that cannot be answered through conceptual analysis alone.
The paper draws on roughly 90 sources across media ecology, learning sciences, critical algorithm studies, postdigital education, and educational policy. Use search to filter.