The study I choose examines the effectiveness of mobile adaptive personalization of mobile news channels by researching and answering the following three questions:

1. To what extent can adaptive automatic personalization produce better results over time by learning more and more about the user?
2. Is automatic adaptive personalization better than customization?
3. How does the information about the users’ social networks’ preferences influence the performance of the personalization?

The researchers set up an adaptive personalization system, that combines the data on the behavior of the user on the site and information about interests of the users’ social network to offer better and more relevant recommendations about news articles on mobile devices. This area is increasingly important to research, as we tend to read the news more and more on mobile devices that impose several constraints such as small size of the display, limited wireless connections and reduced attention spans of the readers. It is crucial to reduce the information overflow and only display the relevant news.

Fortunately, it is fairly easy to customize and personalize these feeds without significant additional costs (compared to newspapers) thanks to IS technology. Therefore, the key question becomes: How should these products be adapted? Currently, most of these sites allow the users to self-customize their feed by choosing topics and categories that they are interested in. However, even if the person is interested in sports, he/she might prefers to read about the local sports news as opposed to other regions’.

Other than that, customers’ preferences change over time and it requires effort to always change their preference settings according to their new interests. Therefore, automated adaptive personalization combined with the customer behavior data can work better. The adapted algorithm doesn’t require proactive effort on part of the user, filters the news based on the individuals’ and his/her peers’ behavior, and adapts to the changes in reading preferences. Whether or not the user decides to read one of the articles becomes additional information to use for future recommendations.

Additionally, the algorithm evaluates the users’ social networks’ reading history and in case the users’ preference is ambiguous, it relies on the peers’ decisions and influence.  The privacy concerns of the users is eliminated in the system, because it uses the social network data anonymously without the users being aware of the source of the recommendation.

The following result were found:

1. As expected, the performance of the algorithm improved with the time, increasing the number of read personalized articles, as well as the time spent on reading the articles.
2. The adaptive algorithm significantly outperformed the self-customization method suggesting that users are much more selective than what the offered categories indicate.
3. Finally, using information of social networks increased performance of the model by 7,8%. However, only 8% of the articles were presented based on social network information, as the algorithm itself offered sufficient personalized results.

The findings suggest that adaptive personalization combined with social network data can be a successful personalization approach.

I also found an application calles Delvv, which works with a similar approach.

Source:

Chung, Tuck Siong, Michel Wedel, and Roland T. Rust. “Adaptive personalization using social networks.” Journal of the Academy of Marketing Science 44.1 (2016): 66-87.