Developing fair user recommendation for decentralized SNSs (Mastodon, Pleroma, etc.)
I develop a fair user recommendation system for decentralized SNSs.
Decentralized SNS is an SNS consisting of a network with multiple servers (instances) connected. Currently, each implementation of decentralized SNSs does not have the function of user recommendation (who to follow), or only recommends the users who belong to the same server. In order to discover unknown users across multiple servers, third-party user recommendation systems are required. User Matching for GNU social, Mastodon, Pleroma, and microblog.pub solves this task and provides the user recommendation across multiple servers.
However, many of the user recommendation systems which are designed naively, recommend only already popular users. This leads to oligopoly and centralization of users and is not good for our user community. Now I define fair user recommendation as the one that actively promotes users who are not well known. In contrast, Unfair user recommendation is defined as the one that brings oligopoly and centralization of users. I am conducting activities to evaluate qualitatively and quantitatively the fairness of the user recommendation systems. For details, see this slide (The Recommendation Fairness) or this report (Decentralization by design: How to get back the internet of celebrities into the 99 percents' hands).
Funding is needed to develop and operate my user recommendation system for distributed SNSs.
At the moment, the following decentralized SNSs are supported.
hakabahitoyo owns the following accounts on other platforms:
hakabahitoyo joined 7 months ago.