Introduction. The objectives of this research are: 1) to develop an algorithm for automating the social media users’ arguments regarding self-preserving behavior (motivation to continue or quit smoking); and 2) to structure the reasons to continue or quit smoking among the Russian-speaking users, based on testing the developed automation algorithm to rest the future arguments of demographic policy measures.

Materials and Methods. The algorithm to classify social media users’ arguments in favor of or against smoking cessation was developed using natural language processing methods based on the Conversational RuBERT neural model. The authors compiled a dataset of over 40,000 Russian-language comments posted on YouTube for model training.

Results. A system of opinions among Russian-speaking YouTube users regarding self-preservation behavior was developed using thematic analysis of demographic content of search engines (concerning smoking cessation). According to our findings, health preservation is the predominant reason in arguments supporting smoking quitting, outweighing financial considerations. Additionally, the desire to avoid weight
gain was identified as a reason some users choose not to quit smoking, although this factor is not a primary concern. On average, class prediction accuracy exceeds 85 %, indicating a high level of results reliability.

Conclusions. The algorithm developed by the authors for automating the arguments related to smoking cessation provides real-time insights into the factors most strongly deterring Russians from smoking cessation (whether health concerns or cost of cigarettes) and the prevalence of certain myths about smoking cessation in the Russian society. The obtained data can be used to better tailor demographic policy measures aimed at reducing smoking prevalence in Russia, potentially leading to quicker and more effective outcomes.

Keywords: self-preserving behaviour, tobacco smoking, neural network methods, digital demography, machine learning, social networks, Russia