How recommendation agents can restore trust lost through promoting sponsored products

A review of the article “Effects of Recommendation Neutrality and Sponsorship Disclosure on Trust vs. Distrust in Online Recommendation Agents: Moderating Role of Explanations for Organic Recommendations” by Wang, Xu, and Wang (2018)

Ever wondered why a pair of 5-inch stilettos, completely out of your price range, shows up as the first product option when searching for a pair of trainers online? Much of the answer lies in how recommendation agents (RAs) – essentially the algorithm tool that sits on many online stores’ and aggregator’s websites – reflect neutral or bias outcomes. Neutral RAs will display outcomes based solely on the user’s requirements. However, biased RAs manipulate the outcomes to promote sponsored products, irrespective of its relevance to the user’s requirements. In the adjacent example, a user searching for “rainbow socks” is first presented with a sponsored box-set of socks and, below that, the product aligned to his/her requirements.

Amazon search of rainbow socks

The Study

This study explored the extent to which consumers perceive their psychological contract – a “contract” that online users establish to guide their trust and interaction with websites – to have been violated through the manipulation of search outcomes to favour sponsored products. The outcome of this violation is a loss of trust or even creating distrust in the recommendation agent. A second study looked at how the disclosure of sponsorships or the display of explanations can restore lost trust or reverse distrust that has settled in.  Data was collected by means of a lab-experiment during which the researchers manipulated the order of search outcomes to simulate bias and neutral RAs and then surveyed participants on trust and distrust.


  • Neutral vs bias: participant’s had higher levels of trust and lower levels of distrust in neutral RAs when compared to bias RAs.
  • Sponsorship disclosure: participants had higher trust in a biased RA with sponsorship disclosure than a biased RA without sponsorship disclosure. However, sponsorship disclosure did not lower distrust.
  • Explanations: participants had higher trust in a biased RA with explanations than a bias RA without explanations. Again, distrust was not resolved by displaying explanations.

The same study was completed in Hong Kong which confirmed the findings from the USA and that the findings are applicable cross-borders.

Practical applications

We see the roles of recommendation agents and sponsored advertisements come to life on social media platforms such as Instagram. As a celebrity with more than 120 million followers, Kylie Jenner is an influencer. Her followers would pay attention to the products she uses. However, based on her Instagram posts, her followers do not know whether or not a product is sponsored. In the example below, we do not know whether Lyfe Tea is a recommendation coming from Jenner or a third-party sponsor. Based on the findings of the study, individuals are less likely to trust this product promoted by Jenner, making her post a potential example of bias without sponsorship disclosure.

Kylie Jenner’s non-disclosed sponsored ad

Conversely, there are numerous examples of paid sponsorships on Instagram. Below, we see a post by a social media influencer, where it clearly states that the post was paid for by Volvo. In a snap survey in a recent Masters-level class at RSM, more individuals trusted the below social media influencer over Kylie Jenner because of the disclosure that the post was paid for by Volvo. By seeing that this is a sponsored post, individuals know up front that it is biased. 

Sponsored Volvo Ad

Strengths of the study

The study researches a relevant topic, due to the popularity of e-commerce, and takes a holistic approach in the experiment design. The researchers clearly identify the factors associated with trust, including biased and non-biased sponsorship disclosure, and neutral recommendation agents. The experiment itself focuses on purchasing a camera within an ecosystem designed by the researchers, where they can better manage participants, as opposed to having participants go on a third-party e-commerce website.

Weaknesses of the study

The study doesn’t really reflect consumer decisions – consumers make purchase decisions online based on a tradeoff between perceived benefit, perceived risks and trust(Kim, Ferrin, & Rao, 2008). Even when their trust has been violated, they may still proceed based on their perception that the benefits exceed the risks or distrust they have.

It may also not be appropriate to generalize outcomes of the study which is derived using students as data subjects. Hanel and Vione (2016) found that, when testing personal or attitudinal variables, such generalizing is problematic as students vary randomly from the general public.


The fact that individuals would have higher levels of trust in neutral RAs over biased sponsored-disclosed RAs indicates that individuals are more likely to trust recommendations that are sincere, over paid-for advertisements by a company where the motive for recommending the product may be for financial gain. In the digital age, we may not know who our peers are online, but there is a level of trust established with the majority or popular opinions. 

In regards to motives for consumer-focused businesses, this is an opportunity to create online communities for products. Glossier, a US-based cosmetics startup company, began as a platform where users share reviews on different beauty products. This evolved into Glossier-lined products, which were developed through the reviews on their platform. Although their business model has evolved to be more commercial-focused, the platform still exists. The members of the Glossier community are neutral RAs and have been integral to the development of the company. 

Ethical implications should be considered in the role of recommendation agents. In the example of the Fyre Festival, concert organizers paid social media influencers to promote that they will be attending the event, creating hype that is seemingly organic. While these social media organizers were paid, they did not disclose this fact to their fans; influencers were promoting a concept to the masses without knowing much about the event themselves. Eventually Fyre Festival fell apart, and some individuals believe that the social media influencers are to blame. In fact, the improper use of influencers is now the subject of a $100m class-action claim against the organizers. Thus, having sponsorship disclosure is also a matter of ethics. While this can be detrimental to a business, the practice better informs consumers.

Fyre Festival marketing


Hanel, P. H., & Vione, K. C. (2016). Do student samples provide an accurate estimate of the general public? PloS one, 11(12), e0168354. 

Kim, D. J., Ferrin, D. L., & Rao, H. R. (2008). A trust-based consumer decision-making model in electronic commerce: The role of trust, perceived risk, and their antecedents. Decision support systems, 44(2), 544-564. 

Wang, W., Xu, J., & Wang, M. (2018). Effects of Recommendation Neutrality and Sponsorship Disclosure on Trust vs. Distrust in Online Recommendation Agents: Moderating Role of Explanations for Organic Recommendations. Management Science

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s