In psychology, a cognitive map is a mental model we develop of the world around us: how things are connected, where we are positioned relative to other points, and which paths are possible when conditions change.
Edward Tolman coined the term in 1948 and showed that rats in a maze do not simply follow a memorized route, but instead build an internal map that allows them to take shortcuts when a path is blocked. Humans do the same, not only in physical space but also in conceptual space: we build maps of ideas, systems, institutions, and relationships.
Tools such as GPS, search engines, and today #AI are powerful in part because they take over some of this mapping work for us. However, navigation research shows a drawback: when we rely too heavily on step-by-step instructions, our own spatial orientation skills can weaken.
We may still reach our destination, but we no longer truly know where we are.
The same risk applies to thinking and learning with AI: if we always let a model decide what is relevant, how ideas are connected, or what the next step should be, we may gradually lose our own sense of the “terrain.”
It is not about preserving every map. It is about preserving the ability to create one ourselves.
