The objective of the SELECT project is to help Internet users find the most reliable, valuable, important and interesting information quickly and easily, hence reducing information overload. In these ways, SELECT will make a positive contribution to the problem of helping users tailor their information environments to meet their individual needs. The approach adopted in SELECT is to develop a general architecture for information filtering and recommendation systems, and to use this to implement and evaluate different strategies and techniques. In this paper we describe GroupMark, a prototype of a SELECT-based social recommendation tool for the WWW that is based upon shared bookmarks. We focus in particular on how GroupMark seeks to combine content-based and collaborative filtering techniques, and on the user interface issues raised by recommendation tools: i.e., the mechanisms for controlling behaviour and the visualisation of results.