Information Filtering in Social Networks

Seattle.
OK, I walked in a little late to the first AoIR 2011 presentation this morning, by Michele Willson, whose focus is on information filtering. There are different approaches to such filtering: at the user or at the service end, initiated by users or by the system, cognitive or social filtering, and based on knowledge about the user’s interests which may be acquired through a range of different mechanisms.

Different stakeholders in the process, and in developing these processes, will have a range of different agendas and interests – developers have specific algorithms they may wish to explore, funding bodies and sources have specific commercial or other imperatives, users and their friends are interested in particular forms of online activity (content sharing, phatic communication, etc.), and the social network providers overall are interested in increasing participant numbers and boost the stickiness of the platform.

These interests and agendas combine in various forms to determine the specific filtering processes being used. Filtering processes may or may not be publicly documented (on Facebook, it’s possible to see who else ‘liked’ specific information, for example); they may increase the personal relevance of information available to users; but they may also rely predominantly on black-box code or algorithm; and they enable datamining activities. Additionally, they also offer an opportunity for the conscious selection of ‘friends’ in the social network.

Social filtering also depends on the relationships of trust in the network; how these are operationalised also depends on the specific technological setup of the platform (from basic ‘likes’ to filter lists and other mechanisms). There are both visible and invisible layers to these processes – underlying data patterns are often invisible, and how the platform acts on these patterns may also be invisible to the user, for example. All of these elements also interconnect with one another.

Do these processes decrease diversity in visible information; do they enable surveillance; do they boost sterotyping; do they increase the conflation of public and private; do they contribute to the rationalisation of relationships between people; do they act as a black box; or can they also lead to a more informed citizenship; can they be used for actions of resistance as well?