LSE - Small Logo
LSE - Small Logo

Jiangning He

Xiao Fang

Hongyan Liu

Xindan Li 

July 17th, 2019

Using consumer psychology to enhance mobile app recommendation

0 comments | 4 shares

Estimated reading time: 5 minutes

Jiangning He

Xiao Fang

Hongyan Liu

Xindan Li 

July 17th, 2019

Using consumer psychology to enhance mobile app recommendation

0 comments | 4 shares

Estimated reading time: 5 minutes

Mobile apps play an increasingly important role in people’s lives and enrich almost every aspect, from social networking and health monitoring to shopping and entertaining. People now spend more time using mobile devices than PCs to access the internet. According to statistics from Yahoo’s Flurry analytics, 90 per cent of people’s mobile time is spent using apps, which they can easily download from large platforms such as Apple’s app store and Google Play. In March 2019, Apple’s app store offered 1.8 million mobile apps to download, and Google Play offered even more, about 2.1 million. This huge number of apps available to download on mobile app platforms poses a significant challenge for users trying to locate the apps they desire. It is necessary to develop effective recommendation methods that can narrow down this offer to a handful of apps to meet a user’s needs.

Existing mobile app recommenders mostly predict users’ download behaviours by mining their behavioural patterns, for example recommendations are made so that users who download app A also download app B. However, users’ psychological states and decision process that lead to download behaviours are underexplored. These psychological states are varied. Examples may include conformity, impulsiveness, etc. In this study, we focus on the role of involvement in making effective mobile app recommendations.

Involvement is an internal state of mind a consumer experiences when choosing a product or a service, and the degree of involvement (high or low) is manifested by the amount of time, effort and consideration expended in the process of information search and product acquisition. For example, in purchases of some durable products with high price tags (such as consumer electronics, appliances, and automobiles), consumers are often associated with high involvement and actively search and collect information before purchases; In contrast, while purchasing consumable, low-cost products (such as groceries, instant coffee, and bubble bath soap), consumers are more likely to have low involvement, with few or even no searches and comparisons.

In the case of mobile apps, different apps have different abilities to elicit involvement due to their unique characteristics such as complexity, perceived risk, emotional appeal, and hedonic value. For instance, game apps are more likely to cause high involvement than utility apps, because games have high hedonic value and emotional appeal. In order to effectively locate the apps that a user would be interested in downloading, identifying the user’s internal involvement state and taking it into consideration in the recommendations is indeed helpful.

Even though a user’s involvement with an app download is hidden and unobservable, involvement theory suggests that it is feasible to infer it from the person’s browsing behaviours preceding downloads. According to involvement theory, there is a positive relationship between someone’s degree of involvement and their information search efforts. In the case of mobile apps, information search before an app download is embodied in users’ browsing behaviours on a mobile app platform. A user who has high involvement with a mobile app category may browse many apps in that category to make comparisons before downloading one, while on the other hand, a user who has low involvement with an app category may do little or even no browsing before downloading an app in that category. Therefore, the intensity of browsing behaviours preceding a download is affected by a user’s involvement state with the download, and thus can be used to learn it.

Considering that users’ involvements can be inferred from their browsing behaviours, we propose an innovative recommendation method based on probabilistic graphical model. This method integrates browsing and downloading behaviours to learn users’ hidden involvement states and interests in mobile apps. The information is then incorporated in app recommendations. Using a real-word dataset collected from one of the largest mobile app platforms in China, we conducted extensive experiments to evaluate the performance of our proposed method. The experiment’s results demonstrate the superior performance of our method over several state-of-the-art mobile app recommendation methods.

The results of our experiment demonstrate that considering involvement is crucial to enhance the quality of mobile app recommendations, with important implications for business. First, mobile app platforms can use our method to enhance the performance of mobile app recommendations, which could in turn increase their revenues and profits as well as improve users’ satisfaction with their platforms. Second, our method is ready to be deployed for real-world large-scale mobile app recommendations. When users click on an app on a platform, they will be directed to the page describing the details of the app. Several other mobile apps, namely “related recommendations”, will also be displayed on this page. Our method could be used to generate related recommendations, because it considers recent clicks (i.e., browsing behaviours) to recommend apps. Third, the applicability of our method is not limited to mobile app recommendations; it can also be used to produce recommendations in other domains, for example product recommendations in e-commerce platforms.

Besides making recommendations, our study also provides a data-driven method of discovering users’ involvement from behavioural logs. The proposed probabilistic graphical model could learn the probability that a mobile app category will elicit high or low involvement. Our results show that different mobile app categories can elicit different involvement states. For instance, app categories such as “racing games”, “Mom & Kids”, and “English Learning” very likely cause high involvement whereas app categories such as “Video” and “Navigation Services” are more likely to elicit low involvement. Considering that involvement is prevalently measured through questionnaires, with their unavoidable high cost and user bias, our data-driven method is expected to be a good supplement and can be easily applied to platforms of large-scale users.

We propose an innovative recommendation method premised on involvement theory, integrating users’ download and browsing behaviours to learn involvement and make mobile app recommendations. Broadly, our study is just a typical example of how to infer users’ internal psychological states and consider them to enhance the quality of recommendations. Consumer behaviour theories (e.g., involvement theory in this study) give data scientists new ideas about how to design models to learn hidden psychological states from observable behavioural data. In the future, more psychological factors can be considered, and more research opportunities are about to come.

♣♣♣

Notes:


Jiangning He is an assistant professor at the School of Information Management and Engineering, Shanghai University of Finance and Economics. She received a B.S. from Renmin University of China and a Ph.D. in management science and engineering from Tsinghua University. Her current research interests include personalized recommender systems, social network analytics, consumer psychology and intelligent competitor analysis.

Xiao Fang is a professor of management information systems and JPMorgan Chase fellow at Lerner College of Business & Economics and Institute for Financial Services Analytics, University of Delaware. He studies business analytics, social network analytics, and financial technology with research methods and tools drawn from reference disciplines including management science (e.g., optimisation) and computer science (e.g., machine learning).

Hongyan Liu is a professor at the School of Economics and Management, Tsinghua University. She received her Ph.D. in management science from Tsinghua University, China. Her recent research interests include personalised recommendation, social computing, health analytics, data mining and machine learning.

 

Xindan Li is Zhaoshiliang Chair professor at the School of Management and Engineering, Nanjing University. He also serves as the director of the Institute of Financial Innovation at Nanjing University. His current research interests include behavioural finance, risk management, and financial technology.

 

 

 

About the author

Jiangning He

Jiangning He is an assistant professor at the School of Information Management and Engineering, Shanghai University of Finance and Economics. She received a B.S. from Renmin University of China and a Ph.D. in management science and engineering from Tsinghua University. Her current research interests include personalized recommender systems, social network analytics, consumer psychology and intelligent competitor analysis.

Xiao Fang

Xiao Fang is a professor of management information systems and JPMorgan Chase fellow at Lerner College of Business & Economics and Institute for Financial Services Analytics, University of Delaware. He studies business analytics, social network analytics, and financial technology with research methods and tools drawn from reference disciplines including management science (e.g., optimisation) and computer science (e.g., machine learning).

Hongyan Liu

Hongyan Liu is a professor at the School of Economics and Management, Tsinghua University. She received her Ph.D. in management science from Tsinghua University, China. Her recent research interests include personalised recommendation, social computing, health analytics, data mining and machine learning.

Xindan Li 

Xindan Li is Zhaoshiliang Chair professor at the School of Management and Engineering, Nanjing University. He also serves as the director of the Institute of Financial Innovation at Nanjing University. His current research interests include behavioural finance, risk management, and financial technology.

Posted In: Technology

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.