Mise en Place for Taste: Recipes, Connoisseurship, and Cultural Competence in Generative AI

Jun Li

Keywords: Cultural AI Evaluation, Connoisseurship, Mise en Place for Taste, Interpretive Technologies, LLM-as-Judge

TL;DR: Cultural AI evaluation inherits the cookbook's bet that taste can be written down. We propose mise en place for taste, an evaluation paradigm targeting the conditions for connoisseurship rather than its outputs.

Abstract: Monet did not cook. The recipes in *The Monet Cookbook* were prepared by a household kitchen the painter hosted, sourced for, and appreciated but did not author. We argue this gap is a precise analogy for contemporary cultural AI evaluation, which has inherited a 19th-century *specification thesis*: that cultural competence can be written down and transmitted at scale. The French gastronomic literature from Brillat-Savarin through Babinski records the thesis's structural failure mode: connoisseurship (trained, indexical, relational, untransferable judgment) cannot be specified, only deferred to a competent agent at execution time. The cookbook genre marks these positions with the phrase *to taste*. We propose taking this primitive seriously and reorganizing evaluation around a paradigm we call *mise en place for taste*: targeting the conditions under which competent judgment can arise, rather than auditing outputs against rubrics. Four concrete proposals follow. By separating cooking from hosting, the argument also addresses the displacement risk that better cultural AI poses to working artists.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 27