Satisfaction with Behavior Trajectory Security Protection in Social Media Intelligent Recommendations

2.1 Intelligent Recommendation and Trajectory Information

Intelligent mobile terminal devices are becoming increasingly sophisticated in their functions, and operating systems are constantly improving their ecosystems, allowing for a wider range of collectable data. In addition to having access to user privacy data such as their locations, posted posts, and social relationships, social media platforms capture, mine and analyze historical user behavior data. This is done through the use of machine learning tools and algorithms to provide users with personalized recommendations and search predictions that optimize recommendation algorithms (Ge and Persia 2019; Lai and Zheng 2015). To fulfill the needs of intelligent recommendations, the recommender systems collect data within the application and capture switching and sharing activities across applications. In addition, this also acquires cross-application information including application data list and clipboard usage (CNCERT/CC and CSAC 2021). While privacy-related academic achievements recognize users’ trajectory information in the mobile environment, most current research (Al-Hussaeni, Fung, and Cheung 2014; Dai et al. 2018; Wang et al. 2021) primarily focuses on the physical location aspect of such data, with little attention given to the privacy implications of trajectory data related to digital interactions. In fact, each virtual interaction in the context of intelligent recommendation can generate vast amounts of data encompassing space, time and behavior trajectory, underscoring the critical need for privacy protection.

2.2 Privacy and Security Protection on Social Media

Research on privacy and security protection for social media users has explored several aspects. Regarding the impact of security and privacy protection on user behavior, it concerns privacy paradox in the engagement of consumers with social media and their behaviors related to online privacy protection (Mosteller and Poddar 2017), how trust, privacy and security affect individuals’ utilization of the internet for social media and online shopping (Alshare, Moqbel, and Al-Garni 2019), and the effects of regulatory focus and default on the preference settings of social media users (Cho, Roh, and Park 2019). Regarding the impact of user behavior on security and privacy protection, it concerns the impact of user behaviors on privacy and security threats on social media (Cengiz, Kalem, and Boluk 2022), and the influence of behavior on the security of privacy information on social media (Liu et al. 2022). Regarding countermeasures for security protection, it concerns privacy risks and possible threats on social media, methodology for privacy issues, and trust management (Kumar, Saravanakumar and Deepa 2016). Regarding users’ confidence in security protection and evaluation criteria for security protection, it concerns how to evaluate the level of user confidence in the security and privacy of social media (Ranogajec and Badurina 2021) as well as how to develop a cross-cultural framework to enhance user privacy protection on social media (Ur and Wang 2013).

2.3 Social Media User Satisfaction

Researchers have investigated social media user satisfaction. Yang et al. (2020) used structural equation modeling to analyze the effect of presentation, content, and utility on the satisfaction levels of users who interacted with the official social media platforms of Chinese government. Pang (2021) conducted an online survey of WeChat users and found that hedonic and utilitarian values positively and significantly influence user attitudes and satisfaction of the platform, with satisfaction having a significant impact on eWOM participation. Additionally, utilitarian values indirectly impact eWOM participation through satisfaction. Khan et al. (2023) explored user satisfaction with social media using the Expectancy-Confirmation Model and found that privacy cynicism substantially impacts user satisfaction. Other research explored user satisfaction in different areas within social media, including tourism (Balbi, Misuraca, and Scepi 2018; Martínez-Navalón, Gelashvili, and Saura 2020; Narangajavana Kaosiri et al. 2019), logistics (Alshehri and O’Keefe 2019), and daily life (Uram and Skalski 2022).

3.1 Research Method

The grounded theory method could derive theoretical explanations based on data through induction (Glaser, Strauss, and Strutzel 1968; Guetterman et al. 2019). It is a widely used qualitative research method (Timmermans and Tavory 2012) that has been applied in various studies on privacy and security (Foley and Rooney 2018; Gerlach et al. 2019). This study employed grounded theory to investigate the factors that influence the satisfaction of mobile social media users with the security protection of their behavior trajectories in the context of intelligent recommendation. To facilitate the organization and analysis of textual data and to fully record operational traces, the NVivo software was used for coding (Maher et al. 2018).

3.2 Design of Interview Outline

A semi-structured approach was used in this study to ensure the effectiveness of the interviews. Prior to conducting the interviews, a preliminary interview outline was formulated based on initial literature review, phenomena observation, and relevant research experience. Pilot interviews were conducted with several social media users including both senior and junior ones and, based on their feedback, the interview outline was improved. The final formal interview outline covers four parts.

The first part is “Introduction” which includes the following paragraph:

There is no right or wrong answers to the interview questions. Please answer truthfully based on your personal experiences and thoughts. All information gathered during the interview will be kept strictly confidential and only used for academic research purposes. Your personal information will not be revealed, and the research team will abide by confidentiality regulations and be willing to bear corresponding legal responsibilities. Ethical considerations will be taken into account throughout the research process. The interview will be audio recorded, as this is necessary to meet research requirements.

The second part is “Definition of terms” where several terms such as mobile social media, behavior trajectory, and intelligent recommendation were defined and illustrated. The third part is “General information of interviewees” where demographic information like gender, age, education level, major, location, occupation, years of internet use, and etc. were collected.

The fourth part is “Main questions” which includes the following ten main questions:

1. Do you use mobile social media? If so, please give examples of mobile social media platforms that you have used.

2. Do you understand intelligent recommendation? Please explain your understanding of intelligent recommender systems for mobile social media.

3. To improve the effectiveness of intelligent recommendation content, do you know whether mobile social media intelligent recommender systems collect personal information or privacy? If so, please give examples.

4. Do you think that intelligent recommender systems will track your browsing, searching, liking, reading, or jumping between applications within mobile social media applications? Please provide a detailed explanation.

5. Do you think that mobile social media intelligent recommender systems will track your location and other trajectories? Please provide a detailed explanation.

6. Regarding the behavior trajectories being recorded, do you feel concerned? Please provide a detailed explanation of your thought process and the reasons behind it.

7. Under what circumstances do you feel that your behavior trajectory is secure? Please provide a detailed explanation of the process and the reasons behind it.

8. Please rate or evaluate the satisfaction with the security protection of mobile social media users’ behavior trajectory in the intelligent recommendation context. Please provide a detailed description of the scoring points, lost points, and the reasons, and propose a remedy for the lost points.

9. In terms of behavior trajectory security protection, which App do you trust the most? Please provide a detailed explanation of the reasons and methods that make you trust the App even more.

10. In addition, what other factors do you think would affect the security protection of mobile social media users’ behavior trajectory in the intelligent recommendation context?

3.3 Sample Selection and Data Collection

This study followed a purposive sampling approach and selected participants who could be most informative to the research question (Corbin and Strauss 1990). The research participants were expected to have experience with mobile social media, intelligent recommender systems, and behavior trajectory security protection in order to provide substantial information for this research. Preferably, participants with experience, a robust background, and the capacity to understand and provide relevant information for the interview were selected. It was intended that they possessed higher information literacy, a wealth of knowledge reserves, clarity in expression, and a fitting background to the research. Participants must satisfy the interview requirements on certain aspects, such as time and energy, and consent to a complete recording of the interview.

The interviews were conducted over the course of five months to collect the necessary data for this research, with the interviewing process stopped after thorough verification of theoretical saturation. Ultimately, 22 interviewees were selected as the research sample for this research, which comprised 17 online and five offline sessions, with all interviewees consenting to the recording of the interviews. During the formal interviews, the questions were adjusted based on the respondents’ specific answers. Interviewers asked follow-up questions for responses that were incomplete or unclear and for topics that required greater depth. The recordings of the interviews were first transcribed once the interviews were complete, with the transcripts then manually compared to the recordings. Subsequently, the interviewees checked the transcripts for errors; in the absence of any errors, the transcripts were used as subsequent research materials. The 22 research materials were named as A to V, and stored as 22 files.

4.1 Open Coding

This study analyzed original interview data sentence by sentence. In the process of open coding, “indigenous concepts” were elicited from the original statements as far as possible, while duplicate concepts were coded only once. In the end, 271 initial concepts were obtained, with examples of how initial concepts were elicited from original statements shown in Table 1.

This study carefully clustered and categorized initial concepts, resulting in 26 subcategories. The subcategories and initial concepts are shown in Table 2, which reflects data support during the identification and elicitation of subcategories.

Table 3 illustrates the connotation and related literature regarding the 26 subcategories, demonstrating literature support during the identification of subcategories.

4.2 Axial Coding

During axial coding, the 26 subcategories were reorganized and clustered based on prior literature, resulting in 10 categories, as presented in Table 4.

4.3 Selective Coding

During selective coding, after careful and repeated comparison, the relationships among categories emerged as revolving around satisfaction with behavior trajectory security protection. Consequently, satisfaction with behavior trajectory security protection was identified as the core category. Table 5 displays the typical relationship between categories. Figure 1 shows a theoretical model regarding the influencing mechanism of satisfaction with behavior trajectory security protection in the context of intelligent recommendation.

Figure 1:

Theoretical model of influencing mechanism of satisfaction with behavior trajectory security protection in the context of intelligent recommendation.

4.4 Test of Theoretical Saturation

When no new categories and relationships emerge during coding, theoretical saturation is likely to be achieved (Francis et al. 2010). This study followed the theoretical saturation principle for coding and sample selection. After analyzing the seventeenth sample, no new categories or relationships were identified. To ensure theoretical saturation, an additional five interviews were conducted and analyzed, and no new categories or relationships were identified either. Consequently, it can be suggested that the theoretical model developed in Figure 1 achieved theoretical saturation.

5.1 Discussion of Results

From Figure 1, it can be seen that security protection information quality, security protection system quality, and security protection service quality have direct impacts on satisfaction with behavior trajectory security protection. This is consistent with the information system success model which suggests that information quality, system quality, service quality, and satisfaction are critical for a successful information system (DeLone and McLean 2003). The results suggest that the success of behavior trajectory security protection is embodied in the success of an information system. Indeed, information quality could potentially measure semantic success of behavior trajectory security protection, system quality could potentially measure technical success of behavior trajectory security protection, and service quality could potentially measure application success of behavior trajectory security protection. Additionally, satisfaction could potentially measure effectiveness of behavior trajectory security protection (Zha et al. 2015).

5.2 Implications for Theory

The results have important theoretical implications. First, this study contributes to the development of the theoretical model of the influencing mechanism of satisfaction with behavior trajectory security protection in the context of intelligent recommendation. Second, much research has been conducted to extend the information system success model, however, behavior trajectory security protection in the context of intelligent recommendation has been overlooked in prior literature; this study usefully extended the information system success model to the field of behavior trajectory security protection. Third, following Waldman (2020) who proposed cognitive biases, this study proposed privacy cognitive biases to summarize users’ privacy rights related beliefs. These beliefs include optimistic biases, extreme expectations, excessive self-confidence, and brand influence, all of which were developed in this study.

5.3 Implications for Practice

The results have important practical implications. First, when security protection information quality is high, the security protection information is updated, complete, understandable, accessible, and accurate. When security protection system quality is high, the security protection system is reliable and easy to use. When security protection service quality is high, the security protection service is empathetic, compliant, and controllable. It is thus recommended that developers and managers of behavior trajectory security protection should bear in mind the importance of the information system success model. Second, in the context of intelligent recommendation, users’ behavior trajectories are usually recorded by the system. When behavior trajectories are looked at, listened to, or recorded against a person’s wishes, his or her privacy is breached (Moreham 2014). It can be seen from Figure 1 that three categories related to privacy were identified, namely, privacy literacy, privacy concerns, and privacy cognitive biases. It is found that privacy literacy directly impacts privacy concerns and privacy cognitive biases, both of which have direct impacts on satisfaction with behavior trajectory security protection. It is thus recommended that users’ privacy should be given top consideration and protected as far as possible.