Is the Internet Impartial? Diversity in Knowledge Recommendation in Online Networks

Saurabh Khanna


We are living through times witnessing an incessant data deluge - the largest proportion of which is being generated on the internet. There is a crucial aspect of this paradigmatic shift that we still do not understand. The information recommendations we receive through the internet are curated based not only on our prior digital histories, but also on those of our fellow netizens through ‘collaborative filtering’. Knowledge accessible to an individual is effectively filtered out to predict what she might like on the basis of reactions by ‘similar individuals’ (often defined by network path lengths between individuals). Network based collaborative filtering forms the core of most global knowledge recommendation algorithms like PinSage. Even though these algorithms are adept at promoting usage and digital uptime, we do not know the implications they hold for preserving the diversity of knowledge in online networks. The advantage of bulk knowledge exchange occurring on the internet is that this phenomenon has not stayed intangible any more. My study intends to assess how modern day recommender systems initiate and mold the flow of knowledge through collaboratively filtered recommendations. I analyze the process and implications of knowledge propagation in a granular self-contained network of 1892 music listeners. My analysis looks to answer two critical questions - i) from an individual’s perspective, how does access to knowledge in an online network shape individual preferences over time?, and ii) from a network perspective, is diversity of knowledge preserved in recommendations generated by online networks, or is some knowledge elevated while the other is marginalized? Understanding the dynamics of information access on the internet bears strong implications for preservation of knowledge diversity and consequently an equitable evolution of society.

← Schedule