Resnick, Paul, et al. “GroupLens: an open architecture for collaborative filtering of netnews.” Proceedings of the 1994 ACM conference on Computer supported cooperative work. ACM, 1994.
This seems like a very underrated/underdog work at the intersection of Information Retrieval and Machine Learning. Published in 1994, I wonder what the authors of this paper thought would be the impact of this concept of collaborative filtering. From 1994 to now in 2019 oh boy!
First, let me talk about a quick summary of the publication. Now, that I think about it, I understand why this is an underrated work, since they talk about this ‘ongoing experimentation.’ The authors propose GroupLens as an open-source system using Collaborative Filtering to predict user interest / rating based on other’s ratings with certain principles in mind, namely:
– openness,
– ease of use,
– compatibility,
– scalability and
– privacy.
Since then, collaborative filtering, or should I say, Reccomender Systems have a come a long way, wether it be recommending shows based on preference (Netflix) or products (Amazon / Walmart) or music (Spotify) or even content (Facebook). The authors said, and I quote, “Right now, people read news articles and react to them, but those reactions are wasted. GroupLens is a first step toward mining this hidden resource.”
I am am certain, that this too, like many other technologies, has it’s fair share of advantages, but I would like to focus on the disadvantages that this creates.
The echo-chamber effect on Social Media Sites is the perfect example of overdone Collaborative Filtering. Especially when calculatingly applied on bipartisan articles which may have real-world national/global impacts. Yes I like a given (blue/red) idea better, but the point of social media presence is that I get acquainted with the other (red/blue) idea as well to be able to make an informed decision.