CONCEPT
Statistical Pragmatic Competence
Statistical pragmatic competence is the phenomenon that Winograd's 1986 framework had no vocabulary for: open-domain linguistic capability achieved through mechanisms that do not require being-in-the-world. Large language models navigate ambiguity, resolve context-dependent references, handle metaphor and irony, and produce practically effective outputs across the entirety of human knowledge—not through formal rules operating on explicit representations (classical AI's failed approach), nor through embodied engagement with reality (genuine understanding), but through statistical patterns encoding the traces of human understanding preserved in billions of text samples. The term names a category error made real—a capability classified as requiring understanding, produced by systems that manifestly lack it.
In The You On AI Field Guide
Winograd's framework distinguished sharply between processing (formal manipulation of representations) and understanding (embodied, situated engagement with the world). The distinction was both philosophical and practical: genuine understanding required being-in-the-world, and without it, open-domain competence would be impossible. The large language models falsified the practical prediction while leaving the philosophical distinction intact. They demonstrate that the territory