Research indexJournal article
Культурная асимметрия генеративного ИИ: календарные паттерны и визуальные репрезентации российской модели
Тихомирова, Душкин
Patria8 May 2026Open access
Культурная асимметрия генеративного ИИ: календарные паттерны и визуальные репрезентации российской модели
Тихомирова, Душкин · Patria · 2026
Abstract
This article presents the results of a year-long empirical study of the Russian generative image model YandexART, revealing a fundamental contradiction between its Russian-language interface and the predominantly Western European visual patterns it generates. The research is based on the analysis of 363 images, categorised across 20 parameters, and reveals the phenomenon of ‘‘calendar consciousness’’ in the model — its capacity to consistently associate specific dates with particular visual imagery, colour schemes, and cultural symbols. The study introduces and justifies the concept of two levels of data markup in generative models: technical (engineering processing) and worldview-based (cultural codification). The findings demonstrate that Western datasets’ cultural markup remains embedded within the system’s architecture despite interface localisation and fine-tuning on Russian-language material. The authors interpret the revealed patterns within the context of platform power and risks of cultural colonisation through algorithms. The study further demonstrates that the identified asymmetry is twofold: the model not only substitutes “Western” patterns for “Russian” ones, but also reduces Russian culture itself to an averaged image, erasing the country’s regional and ethnic diversity — a consequence of the systematic absence of such material in the training data. The research contributes to developing methodologies for cultural auditing of generative systems and formulating cultural sovereignty policies in the age of AI.