> [!cite]- Metadata > 2025-08-20 15:29 > Status: #paper > Tags: `Read Time: 2m 48s` > [!Abstract] [A Framework for Representing Knowledge](https://dspace.mit.edu/handle/1721.1/6089) > This is a partial theory of thinking, combining a number of classical and modern concepts from psychology, linguistics, and AI. Whenever one encounters a new situation (or makes a substantial change in one's viewpoint) he selects from memory a structure called a frame, a remembered framework to be adopted to fit reality by changing details as necessary. ### One-Sentence Summary Minsky introduces the concept of frames—structured data representations that capture stereotypical situations—as a way for AI systems to represent and use knowledge flexibly, bridging perception, memory, and reasoning. ### Concepts Explained for Beginners **Knowledge representation**: How computers store and organize what they "know." **Frames:** Imagine them like mental templates or skeletons that get filled in when we encounter familiar situations. **Slots & values:** Each frame has slots (categories, like “room has: walls, windows, floor”) that can be filled with specific values (e.g., “walls = brick”). **Defaults:** If information is missing, frames provide default assumptions (e.g., if you think of a “classroom,” you assume “desks” are present, even if not explicitly stated). ### Key Beginner-Friendly Ideas **Humans use stereotypes**: Our minds rely on defaults and patterns when we see something new. **Frames store these stereotypes**: They let AI use assumptions, then update when details differ. **Contextual flexibility**: The same frame can be adapted to many variations (e.g., “restaurant frame” covers fast food, fine dining, or a cafeteria). **Bridging memory and perception**: Frames link what we expect with what we observe. ### Key Components & Steps **Frame creation** – Define a stereotypical situation (e.g., “going to a birthday party”). **Slots (attributes)** – Identify elements involved (guests, cake, candles, gifts). **Default values** – Pre-fill expectations (cake has candles, guests bring gifts). **Instantiation** – When faced with a real situation, fill in actual details. **Adaptation** – Adjust defaults when reality differs (maybe there’s no cake at this party). ### Applications **Natural language understanding**: AI uses frames to interpret sentences in context (“He ordered a burger” assumes a restaurant frame). **Computer vision**: Frames help interpret incomplete images by filling in missing details. **Robotics**: Robots use frames to act in environments with incomplete info. **Education software**: Teaching machines can anticipate common student misunderstandings using frames. ### Examples of Use Restaurant frame: - Slots: menu, waiter, customer, food, payment. - Default values: waiter brings menu, customer orders, food is served, bill arrives. Office frame: - Slots: desk, computer, phone, employee. - Default values: employees work at desks, computers used for tasks. ### Related Areas of Study Schemas (Psychology) – Mental structures organizing knowledge (very similar to frames). Scripts (AI & Cognitive Science) – Step-by-step event sequences (e.g., “ordering food”). Ontology in AI – Formal models of categories and relationships. FrameNet & Semantic Web – Computational linguistics projects inspired by frames. ### Thesis – Antithesis – Synthesis Thesis: Knowledge can be represented using structured frames that encode expectations, defaults, and variations. Antithesis: Human knowledge is too dynamic, fuzzy, and context-sensitive to be rigidly captured in frames. Synthesis: Frames work best when combined with probabilistic reasoning, learning, and flexible adaptation—forming the basis of many modern AI systems. --- ### **References** Marvin Minsky (1974). A Framework for Representing Knowledge. MIT-AI Laboratory Memo 306. Schank & Abelson (1977), Scripts, Plans, Goals and Understanding. Barsalou (1992), Frames, Concepts, and Cognitive Science. Fillmore (1982), Frame Semantics.