Discussion

The invention of the automatic recommendation system brought more opportunities and efficiency for both customers and providers. Nevertheless, it also bore large trust issues due to its digital property. It has abandoned the trust between real person contacts. According to Nielsen’s Global Trust Report on Advertising, 92% of consumers believe that suggestions from friends and family are more important than advertising (Nielson, 2012). At the same time, the rapid development of the Internet has also stimulated the rapid generation and rise of Internet celebrities. As supported in hypothesis 1, when a general automatic recommendation system included curators (influencers) into its system, consumer’s willingness to accept the recommendations would be increased, which consisted with the result of previous researches that influencers are considered more credible and knowledgeable (Berger, 2016). Influencers themselves are consumers, users would have a sense of identity with the curators. Simultaneously, curators also share the function of celebrities, such as the halo effect common consumers would not have (Tapinfluence & Nielsen, 2016).

Trust is an important and fundamental element of consumer intention, especially in this era of the universal digital era. Such as intention to repurchase, intention to churn a subscription, or intention to accept the recommendations given by vendors, etc. With the adoption of RS, the relationship between RS and users is a state of dependence and this dependence will entail risk. Trust among two parties becomes very crucial (Chopra and Wallace 2003). The result of the mediation test in hypothesis 2 has shown that trust has a full mediation effect between RSs adoption and user’s willingness to accept, which indicates that willingness of accepting the recommendations only happened due to user’s trust in RSs and this conclusion has consisted with previous researches examined the same field.

Moreover, as inferred in the conceptual model, both the expertise and popularity of the curator have a positive effect on the willingness of accepting the recommendations, which means that both of them are important drivers for users to accept the offered recommendations. Under the deeper analysis of the data, the expertise level of curators has a larger impact on user’s willingness to accept the recommendations than the curator’s popularity. Although the smartphone industry is a relatively unisexual field, but because it is a technology based-product, the professional requirements for its function evaluation are relatively high. The consumer would concern more into the performance of smartphones and therefore it might cause the choice bias in the expertise-based curator from users. It has made the expertise of the curator seems more important for user's trust perception, but it depends on the industry that researches examine.

In marketing field, transparency is usually viewed very broadly as a trust building approach (Bentele & Seidenglanz, 2008; Donaldson & O’Toole, 2000; Sheppard & Sherman, 1998). However, the hypothesis is rejected and the result told that transparency is presenting a negative effect on trust. Similar results also happened in other papers. Such as in the paper of Audrey

However, the hypothesis is rejected and the result told that transparency is presenting a negative effect on trust. Similar results also happened in other papers. Such as in the paper of Audrey Portes, Gilles N’Goala, Anne-Sophine Cases (2020), they segmented transparency into multiple dimensions. They figured out that different dimensions of transparency might bring different effects on consumers' trust in a particular field. Further analysis indicated that there is a potential quadratic U-shaped relationship between transparency and trust, which means that out of certain threshold, transparency would lose its effectiveness in enhancing trust. Apart from that, people have limited rationality. Overly or complex additional information instead of induce reassuring, would generate new uncertainties that can induce negative effects (Lowrey, 1998). For instance, when users know how the recommendation system works, they may start to doubt on the usage of their information.

Additionally, age, gender, and education level differences didn’t demonstrate any significant impact on their trusting intention and the willingness of accepting the recommendations, which was not expected. It might due to the homogeneity of the collected samples.

Implications

This research provides managers with better insights and enables them to understand whether the participation of curators could more effectively increase users’ trust in the automatic recommendation system. So that their acceptance of recommendations from online systems increases. According to hypothesis 2, trust plays an important role in consumers’ intention to actually take the recommendation into their consideration or even buy products from the recommendation. Therefore, vendors, social media, and influencers, all should focus more on how to increase their trustworthiness towards consumers. Inspired by hypotheses 1, 3, several approaches could use to increase consumer perceived trust in RSs.

Firstly, with the rising of the Internet celebrity economy, many brand managers and advertisers are already striving to find excellent influencers to achieve advertising effects. To further develop the strategy with our findings, the managers of all kinds of the platform that uses automatic recommendation systems should try to foster their own curators, which would attract more business chances from product providers. Secondly, hypothesis 3 confirmed that high popularity and high expertise could both enhance users’ intention to trust into the RSs. Therefore, when brand managers premeditate on which curator would be the best choice to promote their products, they should consider these two attributes. In addition, these findings also prove that the different characteristics of curator will also bring different impacts. Brand managers should choose collaborators according to the attributes of their respective companies.

All in all, these contributions could bring benefits and opportunities to both recommendation vendors and merchants who could like to take advantage of the automatic recommendation system.

Limitations

This study entails several limitations that could hamper the validity and representativity of the results. First of all, the research was conducted in an online survey form, and it was almost impossible to control external interference. Interference caused by external factors might affect the experience and the answers to the questionnaire. Besides, most of the surveys are predesigned by the questionnaire designer to answer the scope, making the respondent more limited in answering the survey, which might cause a miss of more detailed and in-depth information. In addition, online surveys constrained the possibility of actual interaction between the respondents and prototyped RSs. It would affect the authenticity of responses or provides a wrong perception that the researcher wanted to get manipulated. At the same time, all measures were distributed in English, which may have caused difficulties for non-native English speakers.

Secondly, the central limit theorem sampling approach only required the minimum number of respondents. At the same time, due to the convenience sampling method, the sample mainly entailed respondents from the university, which consisted mostly of age group 18-34, and had bachelor's degree or master's degree. This may lead to serious deviations in the data. Moreover, as the participants are highly educated, they might realize the intention behind the survey. Therefore, the feasibility of extrapolating the results to the general population was questioned.

Thirdly, due to the time and length of the questionnaire, this research only used trusting intention to represent the larger concept: "trust" which is not comprehensive enough.

Further research

As mentioned in the limitation section, this paper only focused on trusting intention which is not comprehensive enough to present the whole concept “trust”. According to the result of hypothesis 2, trust is a strong mediator towards consumer’s willingness to accept the recommendations have been provided or showed. Moreover. trust is also a very complicated topic, and hard to measure of being a subjective variable. In the future, researcher could examine trust more in-depth with more dimensions. The moderation test was failure, but there still space to improve. Furthermore, there are many more elements that would have impact on the relationship between user’s acceptance on RSs and the RSs itself can be investigated.

Furthermore, lacking of trust does not only induce rejection on accepting the recommendations, also consumer’s privacy concerns. Consumers questions about the usage of the recorded information they offered consciously and even unconsciously, which leads to a unwillingness of providing information, giving trust and consider the recommendations provided. It’s one of the biggest issue in online commerce environment. Researcher could also work on this direction towards RSs for future investigations.

According to the unexpected find out on transparency in this paper and other papers, transparency might have different impacts on people’s trusting intention could be a large topic and interesting that worth to investigate.

In addition, if future investigators are interested in the intervention of Internet celebrity economy in RSs, the attributes of curators also not just limited with their expertise and popularity. It is also noteworthy to look into other dimensions.