Definition
A symbol (in cognitive science) is a representational unit — a pattern that stands for something in the world, in the mind, or in an internal model. In classical symbolic AI symbols are static tokens (Lisp atoms, Prolog terms). In Hofstadter's account they are active — self-sustaining brain patterns that represent and that can trigger other symbols.
Why it matters
How it works
In symbolic AI, symbols are atomic tokens manipulated by rule-based systems. In connectionism, symbols are distributed activation patterns over many units. In Hofstadter's active-symbol picture, a symbol is a self-sustaining pattern that becomes active when triggered, primes related symbols, and influences behavior. Modern transformer embeddings sit between these views: continuous vectors that nonetheless play the functional roles of symbols (substitution, composition, analogy).