Recommender Systems - Stalking or Advancing?

 

How many of you, while browsing on a website, have looked at a particular item and been given suggestions by the site telling you, "You Might Also Like...?" or "People Who Like This Also Like This?" I'm sure most all of us have been there. Perhaps we have made comments about how "creepy" or "weird" it is that advertisers and websites tend to know us on a more individual basis; and while it does appear to be a bit strange that our individual interests are being recorded, recommender systems are being used in many application settings to suggest products, services, and information items to potential consumers (Huang, 2007), thus becoming the ultimate successor in terms of satisfying the user's interests as well as gaining profit. Companies such as Amazon and Netflix have successfully deployed commercial recommender systems and reported increased Web and catalog sales and improved customer loyalty (Huang, 2007). In Chris Anderson's article, "The Long Tail," for example, it is described how Amazon recommendations brought about the popularity of a not-so popular novel. "The online bookseller's software noted patterns in buying behavior and suggested that readers who liked Into Thin Air [a successful novel by Jon Krakauer] would also like Touching the Void, a similar, but not-so successful novel written a decade prior to Krakauer's]” (Anderson, 2004). Recording the patterns of users on Amazon allowed the site to hone in on the individual interests of the users who liked one novel, and suggested to them the other novel – thus giving the less popular novel a great deal of recognition and gaining profit from the buyer who purchased the recommendation. Recommender systems are an "example of an entirely new economic model for the media and entertainment industries" (Anderson, 2004), and at the heart of this new economic model are the algorithms for making recommendations based on various types of input data. Most recommendation algorithms take as input the following three types of data: product attributes, consumer attributes, and previous interactions between consumers and products (e.g., buying, rating, and catalog browsing) (Huang, 2007). 

 

Netflix is an example of an application that utilizes recommender system algorithm aimed towards rating and catalog browsing as Netflix aims to satisfy the individual user, and not necessarily recommend a product that they hope to gain a financial profit from. To recommend movies and television shows to the Netflix user, Netflix merely bases the recommendations off of the ratings that the user provides to them with their five star rating scale. But as we all know, there is inevitably some grey area in a five-star rating system. Perhaps you watch a horrible Netflix movie, so you give it one star. Then after that movie, you watch another, which is ultimately a much better movie than the first, so you give it five stars. What would have happened if you watched the second movie first? Perhaps you would only have given it a three star rating. But can Netflix take these human factors into account? “It's easy to say you should take human factors into account — but how, exactly? How can you use psychology to study people about whom you know nothing except what movies they like” (Ellenberg, 2008)? In my opinion, merging psychology and technology is the next step in further developing recommender systems and advancing technology altogether.

Recommender systems, as described in Cosley’s, “Is Seeing Believing?” article states that they “are one tool that uses people’s opinions about items in an information domain in order to help people make decisions about which other items to consume.” Knowing what a user likes and dislikes, will inevitably result in satisfying the user as the information being presented to them is catered to their individual taste. Ellenberg states that “adding more dimensions” helps to find the “relationships between movies that no film critic could ever have thought of” (Ellenberg, 2008), and Advertising Age agrees: "When performed correctly, such personalization tactics drive higher engagement, increased conversions and greater brand loyalty" (Advertising Age, 2013), which is important in terms of keeping your user happy, but it also shows how individualized technology is becoming. These algorithms used to create the recommender systems are beginning to actually know the individual user beyond the mere demographic, which creates a much larger window of opportunity for people to spend less time figuring out what they want, because technology will ultimately provide them that service.

There has always been a strong emphasis on obtaining demographic information to analyze what media users consume. However, Ellenberg’s article, “This Psychologist Might Outsmart the Math Brains Competing for the Netflix Prize” suggests that demographic information isn’t necessarily what is most important. Analyzing demographics merely groups individuals into categories, but now, it appears that companies and the Internet are becoming more interested in the individual user. With that being said, I am interested in what the future holds in terms of how users will be analyzed. Will the more generalized demographic research and analysis seize to exist because of the stronger emphasis on the individual user? How do you think future technology will be able to obtain more personal information about us?

Works Cited

Advertising Age. (2013, October, 08). How Recommendation Engines Work. Retrieved from

http://adage.com/article/glossary-data-defined/recommendation-engines-work/244583/

Anderson, C. (2004, October). The Long Tail. Retrieved from

http://www.wired.com/wired/archive/12.10/tail.html?pg=1&topic=tail&topic_set=

Cosley, D., Lam, S., Albert, I., Konstan, J., Riedl, J. (2003) Is Seeing Believing? How Recommender Interfaces Affect Users’ Opinions. Retrieved from

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.2.6430&rep=rep1&type=pdf

Ellenberg, J. (2008, February, 25). This Psychologist Might Outsmart the Math Brains Competing for the Netflix Prize. Retrieved from

http://www.wired.com/techbiz/media/magazine/16-03/mf_netflix?currentPage=all

Huang, Z., Zeng, D., Chen, Hsinchun. (2007, September). A Comparative Study of Recommendation Algorithms in E- Commerce Applications. Retrieved from

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.79.8432&rep=rep1&type=pdf