Distributed Cognition
The study of how cognitive activity distributes across people, artefacts and environments rather than residing in any single mind.
Overview
The cognitive unit of analysis is the system, not the individual. A ship navigation team, a cockpit crew, a clinical workflow — each is a cognitive system in which people and artefacts jointly perform tasks no individual could manage alone.
Where extended cognition asks whether individual cognitive systems can extend into the environment, distributed cognition asks a different question: how does a system of people and artefacts accomplish cognitive work that no individual member could perform alone? The theoretical unit is not the mind but the functional system — crew, instruments, charts, procedures and conventions acting as a distributed computational whole.
Edwin Hutchins' Cognition in the Wild (1995) is the landmark study. Observing navigation aboard a US Navy vessel, Hutchins showed that the ship finds its position not through the reasoning of any single navigator but through a distributed process in which information propagates through a sequence of representational states — spoken bearing measurements, plotted positions, alidade angles — as a kind of cognitive pipeline. The computational structure normally attributed to individual reasoning actually describes the system as a whole. Individual crew members manipulate local representations; the ship-level cognition emerges from their coordination.
David Kirsh and Paul Maglio's work on epistemic actions (1994) extends this to individual behaviour: we often act on the world not to achieve a direct practical outcome but to simplify our own cognitive work. Rotating a Tetris piece to check whether it fits, laying out ingredients while cooking, writing a shopping list before leaving — these are actions performed to reduce internal cognitive load by offloading computation to the environment. The environment is recruited as part of the cognitive process, not as a passive medium.
Lucy Suchman's Plans and Situated Actions (1987) provides a complementary critique of cognitivist planning models: actual human action is improvised in relation to local circumstances, not the execution of pre-formed plans. Plans are resources for action, not programmes that determine it. This has deep implications for how AI tools should model and support human work — systems designed around planning models will always misfit the reality of situated practice.
Key Texts
Foundational works in this research tradition.
The founding work. Ethnographic study of navigation aboard a US Navy vessel: the ship as a distributed cognitive system in which information propagates through representational states. Establishes that cognitive processes can and do span persons and artefacts.
Actions performed to simplify thinking, not to directly achieve goals. Tetris players rotate pieces to check fit rather than computing mentally — offloading computation to the environment. Establishes the environment as an active participant in cognition, not a passive backdrop.
How external representations — diagrams, labels, physical layouts — actively participate in cognitive processes rather than merely recording results. External structures are not outputs of cognition but participants in it.
Actual cognition is situated and improvised, not planned in advance. Plans are resources, not programmes. A challenge to AI planning models and a foundation for practice-based design of tools that support work as it actually happens.
The distinction between personal cognition (what an individual does) and system cognition (what person-plus-artefact does). How artefacts redistribute cognitive work — changing not just the difficulty but the structure of thought.
Related Research
Connected areas of inquiry.