Copycat (program)

Copycat is a computer program developed by Douglas Hofstadter and Melanie Mitchell in the 1980s to model the cognitive process of analogy-making in a constrained microdomain. Operating on sequences of letters — such as the classic problem "abc → abd, ijk → ?" — Copycat does not apply pre-encoded rules but discovers patterns through a parallel, stochastic process in which multiple competing interpretations rise and fall in salience. The program embodies Hofstadter's claim that analogy is not a specialized reasoning module but the fundamental operation of cognition: every thought is a mapping between present and past experience, and the quality of the mapping is the quality of the thought.

Copycat's architecture uses a network of codelets — small, competing agents that propose, evaluate, and modify candidate mappings — to simulate the fluid, context-sensitive nature of human conceptual slippage. The program represents an alternative tradition in artificial intelligence that prioritizes cognitive fidelity over performance optimization, modeling how humans think rather than exceeding them through scale or statistical brute force.