# Cognitive Landscapes ## The Distinction A cognitive landscape is a physical system that computes through its own dynamics, not a landscape we compute about or cognitively perceive. Three other uses of the phrase exist, each valuable for its own purposes. Environmental psychology studies how people perceive and respond to landscapes: wayfinding, restorative effects, sense of place. Computational design uses computers to model landscape dynamics: GIS, agent-based simulation, parametric tools. Figurative usage treats landscape as metaphor: "the cognitive landscape of memory." These are not what we mean here. ## The Ontological Claim A *cognitive landscape* is a physical system whose dynamics constitute computation. Not a landscape we compute about. Not a landscape we cognitively perceive. The landscape itself computes. This is an ontological claim, not a methodological one. It asserts that certain physical processes perform information processing through their own dynamics: selecting paths, breaking symmetry, writing memory into structure. Computation is path selection through constraint satisfaction. A watershed computes its drainage network. A fire computes its perimeter. An ant colony computes its foraging trails. These are not metaphors. The shift in perspective is from *landscape as site* to *landscape as process*. A site receives design. A process performs computation. Designers who recognize the landscape as computing can work upstream, adjusting initial conditions so the desired outcome emerges as computed result. ## The Geo-Path as Primitive Geographic information science has oscillated between objects and fields as foundational representations. Both are static. They describe what exists but not how geographic relationships compute themselves. The *geo-path* offers a third primitive: a trajectory through spacetime that constitutes itself through least action. Path finding is path forming. The distinction matters. Optimization problems assume a god's-eye view that surveys all possibilities and selects the best. Physical systems run local algorithms. They do not optimize; they transact. Wheeler-Feynman absorber theory makes the mechanism explicit. A retarded wave propagates forward from the source. An advanced wave propagates backward from the absorber. The path crystallizes where these waves achieve phase-locked standing-wave solutions. The medium performs the computation through bidirectional constraint satisfaction. This is not mysticism; it is physics rewritten to make the computational structure visible. Three conditions determine whether a medium can compute geo-paths: 1. **Coherence**: Phase relationships must be preserved across the medium. Incoherent systems cannot sustain bidirectional transactions. 2. **Criticality**: The system must be poised at phase transition, neither frozen (unable to change) nor chaotic (unable to hold pattern). Self-organized criticality provides the computational substrate. 3. **Phase-preserving coupling**: Bidirectional fields must be able to transact. If forward and backward waves cannot couple, no path forms. The same pattern appears across scales. In physics: photon paths as Wheeler-Feynman transactions. In biology: ant colonies with dual pheromone fields (nest-seeking, food-seeking) that write and read each other's gradients, producing trails that match Fermat's least-time prediction (Oettler et al., *PLOS ONE*, 2013). In geomorphology: watersheds computing drainage networks. In fire: fuel and weather fields coupling with ignition to compute fire perimeters as geodesics through the landscape. ## The Core Question "What properties must a medium possess to compute geo-paths?" This question reframes geographic representation. We ask not "what is the landscape?" but "what can the landscape compute?" The answer depends on whether the medium supports coherence, criticality, and phase-preserving coupling. Some landscapes compute richly. Others are computationally impoverished. Characterizing geographic media by their computational capacity opens new questions: Do watersheds, networks, and social fields share algorithmic structure? What breaks when coherence fails? How does design restore computational capacity to degraded systems? ## Why "Cognitive"? The term *cognitive* does not anthropomorphize. It invokes cognition science, not human cognition. Cognition is information processing that selects paths under constraints. The mind is not in the brain; it is the other way around. The brain is embedded in a cognitive field that extends through body, tools, and environment. Likewise, the cognition of food foraging is not in the ant; the ant is in the cognition. The colony's foraging computation is the larger system within which individual ants participate. Stigmergy provides the clearest illustration. In stigmergic systems, memory is stored in the field, not in the agents. Ants do not remember the trail; the pheromone gradient is the memory. The landscape knows via its structure, not via a mind. The structure is the cognition. This usage aligns with ecological psychology, where cognition is the perception-action loop: not organism OR environment, but the transaction between them. A cognitive landscape is not the environment stripped from the agent. It is a nexus of fields transacting: water and sediment, fuel and fire, forager and pheromone, perception and action, traffic and infrastructure, capital and zoning. Cognition is the transaction, not one side of it. The pheromone field does not cognize separately from ants; the ant-pheromone system cognizes. The watershed does not compute separately from rain; the rain-terrain-channel system computes. Changing the geometry of a landscape changes its least-action flows, and changing those flows changes what the landscape computes. The landscape is not a passive substrate on which cognition acts. The landscape participates in cognition. A cognitive landscape is one where this participation constitutes computation. Design intervenes in the loop: altering geometry, shifting gradients, opening or closing affordances. The designer does not design *for* an agent but designs *the field* that agent and environment share. ## Implications for Design If the landscape computes, design becomes intervention in a running process. The first task is defining the system. A cognitive landscape is not given; it is constituted by the designer's choice of which fields to include. Physical fields (lithosphere, hydrosphere, atmosphere), biological fields (vegetation, fauna, microbiome), and social fields (economic, political, cultural) couple simultaneously, each seeking least-action paths that the others constrain. Landscape urbanism recognized this: the city is not objects but overlapping flow regimes. The designer's initial move is scoping which fields, at which scales, constitute the landscape under design. This is not preliminary to computation; it defines what will compute. The designer does not project form onto a passive site. The designer reads the landscape's computation: identifying gradients, phase states, coupling structures. The designer intervenes upstream, altering initial conditions so the desired form emerges downstream as computed result. Design becomes scripting the landscape's execution. This reframe has practical consequences for fire adaptation, watershed management, urban flow, and ecological restoration. It also has ethical consequences. Working *with* the computation, rather than imposing form against it, positions design as participation in a larger process. The landscape is not material to be shaped but a partner whose dynamics must be understood. ## References - [Geo-Paths as Primitive](https://harvardviz.live/cognitive-landscapes-group/geo-paths-as-primitive.md) - [Step Theory](https://harvardviz.live/cognitive-landscapes-group/step-theory.md) - [Conjugate Variables](https://harvardviz.live/cognitive-landscapes-group/theory/conjugate-variables.md) - [Fire-Adapted Cognitive Landscapes](https://harvardviz.live/cognitive-landscapes-group/fire-adapted-cognitive-landscapes.md) - [Agent-Based Model Abduction](https://harvardviz.live/cognitive-landscapes-group/agent-based-model-abduction.md) - Wheeler, J. A., & Feynman, R. P. (1945). Interaction with the absorber as the mechanism of radiation. *Reviews of Modern Physics*, 17(2-3), 157. - Cramer, J. G. (1986). The transactional interpretation of quantum mechanics. *Reviews of Modern Physics*, 58(3), 647. - Oettler, J., et al. (2013). Fermat's principle of least time predicts refraction of ant trails at substrate borders. *PLOS ONE*, 8(3), e59739.