The Problem with LLM Linguistic Tics: When Automated Language Loses Its Integrity

Summary

This article examines how Large Language Models (LLMs) tend to overuse specific linguistic constructions like "negative parallelism" ("It's not X, it's Y"), and how this has sparked a backlash against automated language production. The author discusses the origins of these linguistic tics in LLMs, what they mean for writing quality, student assessment, and critical thinking. The piece argues that when language becomes quantified and measured by metrics, it loses its integrity and ceases to be good language.