Abstract:

Recommender systems are common tools in e-commerce platforms. In mobile health platforms， course recommender systems assist users in finding relevant exercise courses. Existing literature suggests that recommender systems can effectively cater to user demand， leading to a significant increase in sales for recommended products. However， it remains unclear how recommender systems impact price competition among course providers when users engage in self-searching behavior. Recommendation can be perceived as a form of platform intervention， inspiring users' purchase decision to the recommended products. Additionally， users have autonomous search behaviors. When the recommended courses fail to satisfy users’ needs， they will actively seek alternatives within the same platform or other platforms. Moreover， the influence of different recommended strategies （profit-oriented or user-oriented） and recommender accuracy on platform profits needs further investigation. The design of recommender systems may not solely aim to benefit users， but also to motivate users to buy products that can help platforms to earn more commission fee. Specifically， to achieve higher profits， the recommender systems may misguide users to purchase courses that generate greater revenue for platforms.The impact of different recommendation strategies （user-oriented or profit-oriented） on the profits of a mobile health platform is examined and primarily two key questions are explored： First， how do user-oriented or profit-oriented recommendation strategies affect price competition among exercise course providers and platform profits when users have autonomous search behavior？ Second， how should the platform make optimal decisions between user-oriented and profit-oriented recommendation strategies？To address these research questions， an analytical framework is developed in which two course providers offer substitute fitness courses on a mHealth platform and simultaneously set their prices. The platform charges the providers a percentage commission and recommends only one course to users. Users are not knowledgeable about the fitness courses and need to search individually. In the searching process， optimal stopping rules are adopted to characterize users’ decision-making process before purchasing. The model first considers the baseline scenario in the absence of a recommender system. It analyzes users’ purchase decisions， derives the equilibrium prices of courses， and assesses provider profits. Next， the model assumes that the recommender system exogenously alters users’ searching order. It distinguishes profit-oriented and user-oriented recommendation strategies based on the weights assigned to platform profits and user utility. Furthermore， the model explores the equilibrium prices of courses， provider and platform profits accordingly. Finally， how platforms should select the optimal recommendation strategy in different market environments is discussed.The practical significance is held for mobile health platforms with deployed recommender systems. When the system solely focuses on platform revenue while overlooking providers' strategic responses， or excessively favors either provider profits or user utility， it may potentially harm the platform's profit. Moreover， the factors are identified that determine whether the platform can benefit from the recommender system. These include the recommendation strategy， recommender accuracy， the proportion of random users， and differences in users' course preferences. It has also described how these factors influence the platform's profit.The results show that　（1） When the number of users randomly purchasing exercise courses is relatively high， course providers tend to reduce course price to attract users. When users have diverse preferences for courses and their demand for courses and value assessments also vary significantly， providers have to satisfy different demands through low-price strategies. （2） The high accuracy of the recommender system is more likely to provide users with suitable exercise courses based on their preferences， which helps alleviate price competition among providers. （3） Recommender systems can alter market size but do not necessarily guarantee platform benefits. Platforms can benefit from moderately profit-oriented or user-oriented recommendation strategies. The accuracy of the recommender system contributes to increased platform revenue. （4） When choosing optimal recommendation strategy， platforms should consider commission rates， random user ratio， and the heterogeneity of user health levels within the market environment. Higher commission rates， higher random user ratios， or greater health heterogeneity exert pressure on platforms to adopt a moderately user-oriented recommendation strategy.

Key words: mobile health, recommendation strategies, price competition, search behavior