Suggesting (More) Friends Using the Implicit Social Graph ∗ ∗ Maayan Roth mroth@google.com Tzvika Barenholz tzvikab@google.com Assaf Ben-David abenda@google.com David Deutscher dudo@google.com Guy Flysher guyfl@google.com Avinatan Hassidim avinatan@google.com Ilan Horn ilan@google.com Ari Leichtberg aril@google.com Naty Leiser naty@google.com Yossi Matias yossi@google.com Ron Merom ronme@google.com Google, Inc. Israel R&D Center ABSTRACT Although users of online communication tools rarely catego- rize their contacts into groups such as ”family”, ”co-workers”, or ”jogging buddies”, they nonetheless implicitly cluster con- tacts, by virtue of their interactions with them, forming im- plicit groups. In this paper, we describe the implicit social graph which is formed by users’ interactions with contacts and groups of contacts, and which is distinct from explicit so- cial graphs in which users explicitly add other individuals as their ”friends”. We introduce an interaction-based metric for estimating a user’s affinity to his contacts and groups. We then describe a novel friend suggestion algorithm that uses a user’s implicit social graph to generate a friend group, given a small seed set of contacts which the user has already la- beled as friends. We show experimental results that demon- strate the importance of both implicit group relationships and interaction-based affinity ranking in suggesting friends. Finally, we discuss two applications of the Friend Suggest algorithm that have been released as Gmail features. Categories and Subject Descriptors H.5.3 [Information Systems]: Information Interfaces and Presentation—Group and Organization Interfaces ; I.5.3 [Computing Methodologies]: Pattern Recognition— Clustering General Terms Algorithms, Human Factors ∗ This is an updated version of [16] Keywords Implicit social graph, tie strength, contact group clustering. 1. INTRODUCTION One benefit of many online communication channels over offline methods is that they enable communication among groups of people, rather than restricting communication to be peer-to-peer. Email is just one format that supports group conversations, but there are many others, such as photo- and link-sharing, and collaborative document edit- ing. In fact, group communication is so prevalent that our analysis of the Google Mail email network shows that over 10% of emails are sent to more than one recipient, and over 4% of emails are sent to 5 or more recipients. Within en- terprise domains, group communication is even more criti- cal. An analysis of the email network of Google employees showed that over 40% of emails are sent to more than one recipient, and nearly 10% are sent to 5 or more recipients. As opposed to broadcast-style media, such as blogs 1 and micro-blogging platforms like Twitter 2 , the information com- municated by an individual to a limited group is generally carefully targeted, and may be private. The recipient lists for small-group communications such as emails are selec- tively constructed by the message senders. We have ob- served that users tend to communicate repeatedly with the same groups of contacts. This observation has prompted many online communication platforms to provide their users with tools for creating and saving groups of contacts. Some examples are the Google Mail Contact Manager 3 , or custom friends lists on Facebook 4 . Despite the prevalence of group communication, users do 1 e.g. http://www.blogger.com, http://www.wordpress.com 2 http://www.twitter.com 3 http://mail.google.com/support/bin/answer.py? hl=en&answer=30970 4 http://www.facebook.com/help/#/help.php?page=768