A Framework for Representing Knowledge – Marvin Minsky (1975) [pdf]
Marvin Minsky's 1975 paper introduces "frames" as data structures for representing stereotyped situations in AI. When encountering a new situation, an agent selects a frame and adapts it by modifying details. The work outlines how frames organize knowledge for perception, reasoning, and language understanding.
Background
- Marvin Minsky was a pioneering AI researcher at MIT and co-founder of the MIT AI Lab. This 1975 paper is one of the foundational texts of classical AI and cognitive science.
- Minsky introduces the "frame" — a data structure representing a stereotyped situation (e.g., being in a living room, attending a birthday party). A frame contains "slots" for expected attributes, which can be filled with default values or specific information.
- The paper argues that intelligence requires large stores of commonsense knowledge organized in this framed way, enabling expectation-driven reasoning — not just logical deduction from scratch.
- This work directly inspired "frame-based" expert systems and shaped later developments in object-oriented programming, knowledge representation, and even modern LLMs' use of schemas and templates.
- Written before modern AI winters or deep learning, the paper remains a key reference for understanding how AI researchers once thought (and still think) about representing everyday knowledge.