Distributing Cognition in Collaborative Academic Group Work with AI Support
Cogging
An emerging teaching and learning strategy called “Cogging” is under design to organize informational reading comprehension, shared inquiry, and AI integration in small cooperative knowledge-building groups. Cogging as a patterning of distributed learning opportunities is derived from basic principles outlined in longstanding research in peer-driven collaborative learning with a few unique identifying traits: 1) The task must inherently involve echelons of activity and a wealth of knowledge and information, complex and rich enough to motivate and sustain distributed cognition among small groups; 2) unlike traditional collaborative efforts often intended to yield the same amount and type of knowledge in each mind, working together to complete the cogging task produces different depths and types of knowledge among learners (i.e., all learners learn deeply some but not all of the content in the echelons required to complete the task); and 3) the parameters of the task must be coherent and unified at the learning system and at small group levels, using natural language algorithms as a research tool—the cogs must work together under layers of increasingly expert, well-trained leadership. The remainder of this essay will elaborate on these principles.
*****
“Cog” in this instance is a heuristic, a metaphor serving as a pedagogical anchor. A cog literally is a wheel or bar with a series of projections on its circumference or edge called “teeth,” which transfers physical energy in motion by engaging with teeth on another wheel or bar. An essential component of many physical machines, they are used in systems requiring the transfer of motion or power, such as clocks and watches, bicycles, and car transmissions. The power source—a spring, a battery, human feet, the internal combustion engine—turns a cog which turns another cog and another and so on. They interact with each other through their teeth, which keeps the system working predictably.
A cog metaphorically as in “Cogging” represents a single human mind within a larger learning system or organization of human minds, suggesting that each person's mind contributes to the shared minds in the organization. Just as a cog's teeth must align properly with another cog to function, the metaphor underscores the importance of relationships among individuals in a learning system. An individual cog seems small and replaceable, an insignificant part of a larger entity. The metaphor reminds us that the idea of replacing cogs inside a complex black box of cogs in motion is all but impossible. Even small and seemingly minor cogs are crucial for the system's success. The precision required in the design and use of cogs within a learning system symbolizes the exactitude required in designing expert pedagogy in disciplinary learning.
Three particular mechanical systems without which the civilized and cultured world could freeze up depend on cogs. In printing presses, cogs are fundamental for rolling and pressing functions that apply ink to paper. Our freedom of the press was written into the Bill of Rights because pro-British advocates were confiscating newspaper printing presses and tossing them into the Atlantic Ocean like tea bags. Cogs make up the gear systems of lathes, the machine of machines, to control the movement of the cutting tool and the workpiece. Without cogs there would be no lathes. The lifting mechanisms of many elevators include gear systems that use cogs to smoothly move cargo and people between floors (echelons). Ink to paper to shared governance, cutting tool to workpiece to trains and planes, elevator to floor to corporate headquarters—cogs are put to a wide range of uses in machines and in learning systems including cogging tasks in classrooms.
*****
You haven’t heard of Cogging yet because it hasn’t been fully lathed yet. I’m working on it as a thought experiment to explore controlled and purposeful uses of chatbots in content area classrooms—chatbots as mechanical cogs in cultural and historical activity systems designed with exactitude for human learning. I can offer no evidence from research for the efficacy of any pedagogical design template I’m about to discuss with you. For me, the concept of cogging is intuitively appealing because it connects three central ideas of disciplinary reading: distributed cognition, epistemic reading comprehension, and AI as a research tool. Two of these themes, distributed cognition and uses of AI, need elaboration. Reading to learn as a theme has a long and distinguished history.
Distributed cognition was a hot topic during the 1990s. The cognitive psychologist David Perkins, for example, was interested in how individuals learn and use their expertise to solve problems, pointing to schooling as the system providing access to knowledge and the intelligence required to make knowledge. For Perkins the institution of the school is like a lathe, a master system that hones expert cogs. Perkins was associated with Harvard's Project Zero, still today a thriving resource for contributing to our understanding deep learning and creativity.
As a professor in a secondary subject matter credential program, in my lectures I discussed two of Perkins’ ideas in my content-area reading courses: the fingertip effect and person-plus.. The fingertip effect reminds us that providing a person with a fishing pole, a tool, doesn’t mean they will know how to fish with it. Putting a tool at a person’s fingertips often isn’t enough, and it’s a mistake to assume that it is. It takes the teaching and learning of technical information, understanding of linguistic concepts, and experience in principled inquiry to learn to use AI, for example. "Person-plus" refers to all the external resources an individual has available to support cognitive activity, such as teachers, books, tools, technologies, as well as peers. Now we have natural language algorithms not quite humans, not quite books either.
Cogging as a pedagogical strategy assumes the fingertip effect is true. Students need to learn to use learning tools before they can produce knowledge. Note the subtle shift in the object of pedagogy from directly teaching knowledge to directing teacher tool use (not just strategies, though strategies are a type of tool). Especially when the cog required to move small group learning forward is a human, a peer, person-plus suggests that teachers cannot simply put four cogs into a long term group and expect the group to work like a Swiss watch. Assigning students to use AI in a project without careful instruction is likely to be ineffective or worse as well—just as assigning isolated students to read a chapter in a printed textbook is likely not to produce optimal results without careful instruction and preparation for individual cogging. Students require task-specific cogging preparation and clear expectations from the expert professional in the room, the chief cog in the learning system, the teacher.
Clarifying how individual cogging is designed to interact with other individual coggers in a meshing of teeth doesn’t mean that students can do it. Person-plus assumes that learners have autonomy and space to actively engage in the work of the group, a cog’s autonomy that differs from a cog working alone. With autonomy in a cogging system comes responsibility. Individual students need to work together with available resources in a collaborative classroom in a content-rich context to support knowledge finding and making when they are cogging alone in unsupervised settings.
*****
Seminal research underpinning the first developed theory of distributed cognition was carried out by Edwin Hutchins, a cognitive anthropologist, and published in his 1995 in a book titled Cognition in the Wild. A stunning, electrifying book in itself, its author draws on his own ethnographic research to explore the nature of human cognition in naturalistic settings. The wild in this case is an assault military vessel in the open ocean, an aircraft carrier.
The book's central case study is the navigation practices of US Navy personnel aboard the USS Palau. I have vivid, recurring images from reading this book about the dangerously delicate, intricate maneuvers distributed across echelons of sailors preparing this massive hunk of machinery to make the slow, deliberate turn into the narrow San Diego Harbor without taking out chunks of the harbor itself. One false move and lights out San Diego Harbor. (Click here for a 10-page essay free online written by Hutchins discussing his theory.) Hutchins wrote the following words regarding how the idea for the book started:
“The seed from which this book grew was planted in November 1980, when I spent most of a day on the navigation bridge of a U.S. Navy ship as it worked its way in from the open North Pacific, through the Straits of Juan de Fuca, and down Puget Sound to Seattle. I was aboard the ship to study what the operators of its steam propulsion plant knew and how they went about knowing it. I had spent most of the preceding week down in the bowels of the ship, observing engineering operations and talking to the boiler technicians and machinist's mates who inhabited that hot, wet, noisy tangle of boilers, pumps, and pipes called the engineering spaces.”
Hutchins proposed that problem-solving and learning often involve a network of distributed interactions between agents and tools. Shared knowledge, specialized knowledge, and distributed procedures are the cogs enabling separate and individual acts carried out alone to slowly, carefully turn the aircraft carrier in the space of 1.5 miles, the equivalent of on a dime. The cognitive processes of the ship's crew are shaped both by the social environment, such as the hierarchical structure of the Navy, and by the material environment, including the ship and its equipment. There is no abstract, automatic computational process to bring the ship to harbor, no automatic pilot. Cognition is embodied, situated, context-dependent—human in the moment in a person plus task environment.
The expertise of the ship's crew is a product of learning and practice within the system. Effective performance on ships as in classrooms depends on both individual skills and the ability to operate within a complex socio-technical ecosystem. Cultural practices influence cognition by habituating ways of organizing and communicating thoughts and activities, standard procedures, rituals, and protocols that facilitate coordinated action. Tools like compasses, charts, and computers, also shape cognitive processes. Artifacts carry or produce meaning and require specific skills (specific cogs) to be used effectively within the distributed cognitive system. The collaborative effort of crew members, each contributing different pieces of information and expertise, in the end brings the aircraft carrier safely to shore—or not.
Implications from Hutchins’ theory in the context of human-computer interaction (HCI) seem obvious in this era of simulated verbal intelligence, but the teaching problem is profound, posing a steep professional learning curve, integrating computers capable of simulated human linguistic cognition as cogs in a learning system with humans at the helm.
The functions of AI cut against the grain of American school culture and history where minds are not typically viewed as cogs in a learning system; where learning is typically individual, not distributed; and work that counts is almost always your own, not carried out in concert with peers. One thing wicked problems in education seem to share is this: They are often equal parts technical and material, cultural and political. Hutchins embraced ethnography, studying users in their natural contexts to understand their interactions with technology. This methodology will be crucial in designing strategies like Cogging for seeding local HCI experiences in classrooms.
I’m working on a Cogging design template and an example Cogging project involving distributed learning about the U.S. Constitution, quite a durable and significant cog in modern history. Stay tuned.
