Must a cognitive map be spatial?

Spatial use is central historically, but broader uses describe relationships among concepts or situations.

Is a drawn diagram the same as the mental representation?

No. It is an external expression that may be incomplete or shaped by the task.

Can a cognitive map be wrong?

Yes. It can be partial, distorted or outdated while still guiding behavior.

AI Glossary / Data engineering and architecture / Intermediate

Cognitive map

Definition

A cognitive map is an internal representation of relationships in an environment that helps a person or animal orient, reason or plan. The term is strongly associated with spatial knowledge, such as understanding where places lie relative to one another. It is also used more broadly for representations of how concepts or situations are related.

What it means

A person can know more than a memorized sequence of turns. They may understand that two routes reach the same destination or infer an alternative when one route is blocked. A cognitive-map account describes that relational knowledge. It does not require a literal picture in the mind or a perfectly accurate geometric map. In organisational work, an external cognitive mapping exercise can make someone's perceived relationships visible, but the diagram remains a representation of that person's understanding.

How it works

Experience can supply information about landmarks, connections, distances, sequences and other relations. The learner integrates some of that information into a representation that can support later choices. Researchers infer such representations from behavior and, in some studies, other measurements; they do not simply inspect a complete mental map directly. Computer systems can use maps or graph-like internal models for planning, but similarity in function does not establish that a machine has the same subjective experience or biological mechanism.

Three practical examples

A person navigating a familiar neighborhood takes a new detour because they understand how several streets connect.
A product researcher asks users to sketch where they expect functions to be located, revealing relationships in their understanding of the application.
An AI researcher studies an agent's learned relational representation and tests whether it supports navigation beyond a memorized action sequence.

Think of it this way

Think of a sketch map drawn from memory, showing important places and connections while omitting many details. It can support useful decisions even when distances are distorted. The analogy is close but incomplete: an internal cognitive representation need not be visual, conscious or organized like a paper map, and a drawn sketch may reveal only part of what someone knows.

Worked example

Imagine a small environment with paths A - B, B - C and A - D - C. A traveler usually follows A - B - C. When B - C is blocked, repeating the learned turn sequence fails. If the traveler knows the alternative connections, they can choose A - D - C. This behavior is consistent with a relational map, but one example alone does not prove the internal mechanism: other learned strategies might produce the same choice. A careful experiment would vary starts, goals and blocked paths to distinguish flexible relational knowledge from simpler explanations.

Illustrative case study

Fictional case: Maple Navigation designs an indoor guidance prototype. Users know recognizable landmarks but misunderstand how floors connect. The team interviews participants and compares their route choices with sketches of the building. It revises instructions to reference the landmarks people actually use and tests new routes. The case illustrates using evidence about users' representations, not claiming that a single sketch fully measures a cognitive map.

Common misconceptions and mistakes

Do not assume a cognitive map is a precise miniature of the physical world. People can distort distance, merge locations or rely on selected landmarks. Do not confuse cognitive maps with concept maps, knowledge graphs or fuzzy cognitive maps merely because all can be drawn with nodes and links. These terms refer to different theories or modeling practices. Also, successful navigation by an AI system does not by itself establish human-like cognition; the representation and tested behavior need to be described specifically.

Business use

Businesses can apply the idea when designing navigation, information architecture, training and decision-support tools. Understand the relationships users expect rather than organizing interfaces only around an internal company structure. In strategic workshops, label external maps as elicited viewpoints and compare them with evidence. For AI, test whether a learned representation supports the intended generalization and explain its limits without anthropomorphic claims.

Check your understanding

A user repeatedly visits the same menu path successfully but cannot find the feature after one menu changes. Does that alone demonstrate a flexible map of the application?

Answer: No. The user may have learned a fixed sequence. Testing alternative paths and changed layouts can provide more informative evidence.

Related terms

Concept map
Knowledge representation
Graph
mental model
spatial cognition
wayfinding
model-based planning

Learn next

Ask someone to describe two different routes through a familiar environment and explain how they would respond to a blocked connection. Compare the description with observed behavior. For a digital product, perform the same exercise with alternative paths to a task, while avoiding claims that the exercise reveals the entire internal representation.

Sources

Primary historical research: Tolman, Cognitive Maps in Rats and Men, institutional transcription, https://home.csulb.edu/~cwallis/382/readings/482/tolman.%20cognitive_maps_in_rats_and_men.1948.html ; institutional copy, https://personal.utdallas.edu/~tres/spatial/tolman.pdf . Primary-text search excerpts consulted; broader contemporary interpretation and editorial review remain pending. Examples are explanatory, not evidence establishing a particular cognitive mechanism.

Previous: Cognitive computing Next: Cognitive robotics