دانلود مقاله پیامدهای اقتصادی توصیه های شخصی
چکیده
1. مقدمه
2. مطالب قبلی
3. محیط تحقیق
4. نتایج و بحث
5. مسیر های تحقیقاتی آتی
6. نتیجه گیری
منابع
Abstract
1. Introduction
2. Prior works
3. Research setting
4. Results and discussion
5. Future research directions
6. Conclusion
Acknowledgement
References
چکیده
تأثیر سیستم های توصیه (RS) بر تنوع مصرف شفاف یا به خوبی درک نشده است. مطالعات موجود، چه تجربی و چه نظری، نتایج متناقض و حتی متضاد را نشان می‌دهند، که به عنوان بحث در ادبیات ظاهر می‌شود. در این مقاله، ما تأثیر دو سیستم توصیه‌گر اصلی، فیلتر مشارکتی عصبی و فیلتر محتوای عمیق، بر تنوع فروش را از طریق یک آزمایش میدانی تصادفی بررسی می‌کنیم. نتایج ما توانایی موتورهای توصیه گر را در افزایش یا کاهش تنوع فروش کل تایید می کند. با این وجود، آنها همگن سازی را تقویت کرده و تنوع مصرف در سطح فردی را کاهش می دهند. در نتیجه، تحقیقات ما ظاهراً با یافته‌های قبلی مغایرت دارد و نشان می‌دهد که طراحی RS عامل تعیین‌کننده در همگن کردن یا متنوع کردن فروش محصول است.
توجه! این متن ترجمه ماشینی بوده و توسط مترجمین ای ترجمه، ترجمه نشده است.
Abstract
The impact of recommendation systems (RSs) on the diversity of consumption is not transparent or well understood. Available studies, whether experimental or theoretical, show inconsistent and even opposite results, which manifests as debate in the literature. In this paper, we investigate the impact of two main recommender systems, neural collaborative filtering and deep content filtering, on sales diversity via a randomized field experiment. Our results confirm the capability of recommender engines in increasing or decreasing aggregate sales diversity. Nonetheless, they amplify homogenization and reduce individual-level consumption diversity. In conclusion, our research reconciles seemingly contradict previous findings and illustrates that the design of the RS is the decisive factor in homogenizing or diversifying product sales.
Introduction
Recommender systems are a subgroup of information filtering technologies and are applied to handle the issue of information overload (Schreiner et al., 2019). These systems discover the preferences and interests of users by refining a great amount of dynamically generated data based upon their interactions with items. Subsequently, they predict the willingness of a user to purchase/consume a particular item.
Despite the omnipresence of recommender systems and algorithmic content curation, there are few studies that examine their societal and economic outcomes. Most research efforts attempt to address the technological aspects of recommendation systems and to improve the accuracy of the matchmaking process, whereas papers that seek to elucidate the byproducts of recommender systems at the market level are thin on the ground. In particular, there is a consensus view in the literature that personalized recommendations generate more engagement and more sales (Adomavicius et al., 2018), but the difficulty of gaining access to appropriate research settings and the complexity of personalization algorithms have led to protracted controversy over the distribution of sales and the consumption diversity of users.
Conclusion
Given the current circumstances around the world, online retailers and web-based service providers are gaining in popularity. Recommender systems have been demonstrated to be a powerful tool for these businesses, and their significant value is clear. Therefore, it is of utmost importance to understand the way these intermediaries influence the behaviors of users and how they impact the visibility and sales of different products, from best-selling to niche items. In order to address this issue, we carried out a randomized field experiment employing two most common recommender systems in the context of e-commerce. The analysis of our results shows that both treatments enhance sales; however, collaborative filtering creates concentration bias. In other words, it reinforces the sales of already best-sellers, whereas content-based recommender flattens the distribution of sales and expose users to niche items. Our study reveals that both algorithms decrease individual-level diversity compared to a world without personalization, and homogenization is an inevitable corollary of personalized recommendations. In light of our results, marketing strategists can benefit from a combination of these two matchmaking approaches and find an optimal point that best suits their needs.