ايجاد پرامپت مناسب جهت توليد توضيحات شخصي‌سازي شده در توصيه‌گر فيلم بر اساس ويژگي‌هاي شخصيتي، ابعاد فرهنگي و استراتژي‌هاي متقاعدسازي

Persuasion , Recommender Systems , Large Language Models , pro‎mp‎t Engineering

Creating suitable pro‎mp‎ts for generating personalized descriptions in movie recommender systems based on personality traits, cultural dimensions, an‎d persuasive strategies

مهندسي فناوري اطلاعات

Nowadays, recommender systems play a crucial role in facilitating user decision-making. However, simply offering relevant product suggestions is not sufficient to persuade users. While previous research has addressed the use of persuasive explanations, there remains a significant gap in personalizing these explanations based on users’ psychological an‎d cultural characteristics. To address this gap, this study presents a movie recommender system based on personalized persuasive explanations. In this system, users’ personality types (based on the Neris test, specifically the introversion/extraversion dimension), Hofstede’s cultural dimensions (masculinity/femininity an‎d uncertainty avoidance), an‎d Cialdini’s persuasion strategies (consensus, authority, an‎d liking) an‎d users’ favorite movie genres are taken into account. Two approaches are proposed an‎d developed in this study. The first approach employs a content-based recommendation method based on movie descriptions, directors, an‎d lead actors. The second approach utilizes the BERT language model an‎d Siamese neural networks to prepare a list of recommended films for users. In the next step, personalized persuasive explanations for the selec‎ted movies in these three categories are generated using pro‎mp‎t engineering an‎d advanced language models, including ChatGPT an‎d Claude AI, to eva‎luate how effectively these explanations persuade users to watch the movies. To generate these explanations, a questionnaire aligned with Cialdini’s persuasion strategies, Hofstede’s cultural dimensions, an‎d the Neris personality type test was used, with users scoring certain dimensions in a preliminary questionnaire. The eva‎luation, conducted with the participation of 85 users, showed that men were more influenced by Cialdini’s persuasion strategies (47.6%), whereas women showed a greater inclination toward the masculinity/femininity dimension (39.5%) an‎d uncertainty avoidance (30.2%). The age group of 20–30 years (comprising 60% of participants) was mostly persuaded by Cialdini’s strategies (78.6%) an‎d the introversion/extraversion trait (76.9%). In terms of genre preferences, men showed a greater interest in action, horror, an‎d science fiction genres, while women preferred romance an‎d drama. Additionally, the T-test revealed that persuasive explanations significantly increased users’ average ratings. The average rating for medium-rated movies increased from 3.43 to 4.16, for low-rated movies from 2.97 to 3.58, an‎d for highly rated movies from 3.89 to 4.40.

6

134439