Why AI Education Should Begin with Language and Aristotle

Summary

Hollis Robbins describes a college English course designed to prepare students for AI by starting with language rather than technology. The course uses dictionaries, Aristotle’s Rhetoric and Poetics, and the idea of language as a shared corpus to help students understand words as choices made for particular audiences. The author argues that this foundation is especially important for students who mainly encounter language digitally and may have limited experience with dictionaries, sustained reading, or explicit instruction in how language works. Aristotle’s account of ordinary and unusual words, probability, expectation, style, metaphor, and stale expression provides a bridge to large language models. Robbins explains that an LLM is trained on a pretraining corpus, divides text into tokens, assigns probabilities to possible next tokens, and repeatedly selects from that distribution to generate text. The essay compares this process with rhetoric as a form of social next-word prediction, while emphasizing a key difference: humans know whom they are addressing and what that audience expects. Robbins also contrasts this language-centered approach with AI courses that begin with tokens, entropy, or technical definitions. The essay discusses Anthropic’s use of Aristotelian virtue ethics in Claude’s Constitution, arguing that Aristotle’s treatment of rhetoric may offer a more direct model for tone, proportion, avoiding stock phrases, and adapting language to context.