Recommendations through Analytics: The Flip Side
Recommendations offered through analytics, on the basis of one’s own browsing history is a feature you must have encountered on sites like YouTube, Amazon, Flipkart and IMDb. What this feature basically does is: show up a list of books/ movies you might like depending on the particular book(s)/ movie(s)/ song you have been looking up, or merely a list of books/ movies which were also looked up or liked by other visitors who had earlier looked up the particular piece you were looking for.
Whilst this is a wonderful and helpful feature in many aspects, there is a major drawback to this system. As the analytical tool working in the background works on historical data, one is typically caught in a closed circle of choices. Say, you are searching for a thriller title. Then the list of your recommendations will contain a few more thriller titles. I have found the genre, lead players and directors/ authors to be some of the typical pivots. Now, if you decide to look these up too, you will realize you are often caught in the same group. Most of the titles recommended in the first instance keep recurring in the recommendations for each other, with maybe a few new titles thrown in now and then. This is quite delightful in the short run. But then I discovered to my horror that I am caught in the same universe, traversing my own footsteps back and forth. What this well-meaning tool had done after a point was: restricted me to a set thus limiting my options of venturing into or trying out a completely different set of things, to which there is no link. It is almost like getting too caught in your own social circle and never getting to know the world beyond. This is fine if you feel there is a still lot more to be found out in the space you are in, but bad if you want to explore out too every now and then.
My recommendations 🙂 to come out of this vicious circle:
- In a while, browse without logging into your account
- Then start browsing from a different section on the homepage, or a title which only mildly interests you. It could be your entry to an entirely different rabbit-hole !
It’s an remarkable article in support of all the
online users; they will get advantage from it I am sure.
Avik,
Recommender systems take care of the issue you have mentioned. They will map an user (e.g. you) with a general profile and would recommend the item the generic profile has and you don’t. Basically it’s a failure if it recommends something which you have already seen or rejected!!
Arijit da
Thanks for your comment. I stand corrected!
Perhaps this is a problem one faces when browsing anonymously without logging in. Is it?