Abstract
The article examines changes in the formation of Polish language norms in the context of the widespread use of artificial intelligence technologies, including autocorrect, text prediction, and large language models (LLMs). It introduces the concept of algorithmic normativity into Polish linguistic reflection, defining it as the capacity of computer systems to impose specific language patterns on users through mechanisms of suggestion, correction, and text generation. The article situates this concept within the tradition of Polish normative linguistics and proposes to extend the classical model of norms to include an algorithmic dimension. The empirical section presents the findings of a pilot study comprising two complementary components: an analysis of autocorrect behaviour and LLMs with respect to forms exhibiting normative variation in Polish, and a survey of users’ attitudes towards algorithmic corrections. The results reveal a systematic tendency for algorithms to favour more liberal treatment of linguistic variation, as well as ambivalent attitudes exhibited by language users, who simultaneously declare their trust in
AI-generated corrections and their attachment to traditional institutional norms.