Role of Artificial Intelligence in Tobacco Control: A Scoping Review v1

Background- In the recent years, artificial intelligence (AI) is marking significant impact in fields such as psychology,medicine, public health, etc. and researchers are using AI tools to tackle complex issues in tobacco research. In this scoping review, we aimed tosummarise the published evidences on role of AI interventions in relation to tobacco cessation, motivation, relapse, cravings and identification of promoting factors. METHOD- We aimed to summarise the published evidences on role of AI interventions in relation to tobacco cessation, motivation, relapse, cravings and identification of promoting factors. For this scoping review, we employed the PCC (Population, Concept & Context) framework as per the PRISMA-ScR guidelines.We explored literature using the following four databases- PubMed, Google Scholar, The Cochrane Review& DOAJ. Articles were included if they were peer-reviewed, published in English between 2019 & 2024 and reported the use of AI techniques like conversational agents(CA), machine learning(ML),natural language processing(NLP)etc. in tobacco use. We excluded studies done on individuals using vaping or e-cigarettes. The extracted data included title, author, year of publication, type of study, sample size, socio-demographic characteristics, study strategy, type, name, frequency & limitation of AI intervention and outcome of study. Results- The initial search yielded 1790 articles which were uploaded to Rayyan. Afterremoving duplicates (n=83), the remaining articles were screened and the relevant 42 articles were finalized among which 1 was narrative review and 4 were systematic review and meta-analyses. The studies were diverse in design, strategy, type and platform of AI used. Chatbots (rule based and embodied) and ML were used in some studies to promote abstinence and increase motivation. Some used hybrid AI model for predicting smoking status, lapse risk and treatment outcome while others were iterative developmental study in which they developed algorithm or model of AI and then tested it. Conclusion- While AI hold considerable potential for supporting tobacco cessation, most existing applications relied on rule-based design that provide limited adaptability and resemble standard text-messaging programs. Some studies implemented AI on smallsample which limits their generalizability while in some studies loss to follow up was high. Given the high prevalence of tobacco use in adolescents, we found only 2 such studies. This gap highlights the necessity for further welldesigned studies employing advanced AI driven models.