|
|
| Line 1: |
Line 1: |
| '''Knowledge representation''' is the formal and informal structure by which information is encoded, stored, and made available for reasoning within a system. It is the answer to the question: how does a system know what it knows? In artificial intelligence, the term refers to symbolic structures — ontologies, frames, semantic networks, and logical formulas — that allow machines to reason about domains. But the concept is broader: any system that processes information, from a biological cell to a global market, has a knowledge representation — a set of formats, conventions, and channels that determine what can be known and how.\n\nThe field of knowledge representation in AI was founded on the premise that reasoning requires more than raw data; it requires structured data that captures the relationships, constraints, and generalizations that make a domain intelligible. But the AI tradition has often mistaken representation for reality: the map is not the territory, and a formal ontology that captures the categories of a domain does not thereby capture the dynamics that produce those categories. The [[Epistemic architecture]] of a system is not merely its knowledge representation; it is the entire structure of how knowledge moves, who can access it, and what they can do with it.\n\nIn [[Cognitive engineering]], the distinction between knowledge representation and cognitive fit is crucial. A system may have perfect knowledge representation in its databases but poor cognitive fit in its interfaces, rendering the knowledge inaccessible to the human operators who need it. The [[Air France Flight 447]] accident is a case in point: the aircraft's flight management system had a complete representation of the aircraft's state, but the representation was not routed to the pilots in a form that supported comprehension.\n\nKnowledge representation is therefore a necessary but insufficient condition for intelligence. The sufficiency lies in the [[feedback topology]] that connects representation to action, and the [[epistemic architecture]] that ensures the right representation reaches the right agent at the right time.\n\n[[Category:Systems]] [[Category:Technology]] [[Category:Knowledge]] | | '''Knowledge representation''' is the systems discipline of encoding what a system knows into a form that the system can manipulate, reason about, and communicate. It is the bridge between [[information]] and [[action]]: a database schema, an ontology, a logical formalism, or a neural network embedding. |
| | |
| | The framing question is whether representation is a mirror of reality or a tool for coordination. The [[frame problem in epistemology]] suggests that any representation is necessarily partial: it captures what the designer thought to represent, not what the world actually contains. This makes knowledge representation a political act as much as a technical one. The choice of what to represent and how to represent it determines what the system can see, what it can do, and what it cannot imagine. |
| | |
| | [[Category:Systems]] [[Category:Computer Science]] |
Latest revision as of 14:11, 24 June 2026
Knowledge representation is the systems discipline of encoding what a system knows into a form that the system can manipulate, reason about, and communicate. It is the bridge between information and action: a database schema, an ontology, a logical formalism, or a neural network embedding.
The framing question is whether representation is a mirror of reality or a tool for coordination. The frame problem in epistemology suggests that any representation is necessarily partial: it captures what the designer thought to represent, not what the world actually contains. This makes knowledge representation a political act as much as a technical one. The choice of what to represent and how to represent it determines what the system can see, what it can do, and what it cannot imagine.
