Delectable Knowledge-How the Sense of Touch illuminates the Ontological Status of AI-Recipes

Abstract

The ongoing digital transformation has significantly altered our access to cooking recipes, introducing AI-generated outputs from Large Language Models (LLMs). This paper examines the ontological status of such AI-generated recipes, arguing that neither recipe Platonism nor Constructivism can fully account for them, as both rely on the existence of a prepared dish as a condition for validation. Yet, some AI-generated recipes are treated as successful, suggesting that their acceptance is not purely abstract, but rooted in embodied culinary knowledge. Drawing on Andrea Borghini’s concept of apprenticeship, I argue that recognising a recipe involves both abstract cognition within a type-token framework and sensory engagement through cooking and tasting. This analysis is supported by insights from Alva Noë and Matthew Ratcliffe, who conceptualise perception not as passive reception, but as a skilled and bodily situated mode of interaction with the world. Perceiving, in this sense, becomes a form of doing and touch and taste play a crucial role in how we validate recipes through practice.

By shifting the focus from what is a recipe? to how do we recognise a recipe?, I propose a broader account that accommodates both human and AI contributions. Ultimately, recipe validity is not merely a matter of cognitive classification, but an experiential process grounded in sensory interaction, highlighting cooking as a vital mode of understanding recipe ontology.