SMALL WORDS, DIFFERENT TASKS: RESTRICTION AND QUANTIFICATION IN ENGLISH–UZBEK AI PROMPTS

Abstract

This paper examines how restrictive and quantitative expressions determine the conditions of task completion in English and Uzbek generative AI prompts. Particular attention is given to only, at least, at most and exactly, together with their Uzbek counterparts faqat, kamida, ko‘pi bilan and aynan. Through qualitative analysis of explicitly constructed bilingual examples, the discussion distinguishes restrictions on content, operations and output quantity. It demonstrates how apparently minor changes can alter the set of acceptable responses, especially when a numerical requirement interacts with limited source material. The paper also differentiates formal compliance from evidential adequacy: producing the requested number of examples does not establish that those examples are distinct or supported by the source. A procedure for examining the operator, its domain and its conditions of satisfaction is proposed. The contribution is conceptual and methodological; the examples have not been tested experimentally across models.