Navigating information is an activity that is taking up an increasing amount of people’s lives. Information is imbued in the environment around us, both actual and virtual. We understand the spaces in which we live through our knowledge of specific things and activities and, more frequently, through the application of our general understanding of the systems and approaches that form modern society. Increasing numbers of the new spaces we must navigate are not experienced alone. Therefore it is asking directions of others and interpretation of how others use and behave in a space that aids our navigation of new spaces. This is social navigation: a continuous inquiry into how others navigate a space.
[Please refer to my Bibliography for the sources of quotes.]
Although a comparatively recent development in computing theory and practice social navigation is an important and basic strategy for navigation we use daily. However, on the desktop computer we are rarely asked to employ such navigation strategies. In many cases, when visiting a website the user is not aware of anyone else using the same website. The World Wide Web (WWW) is structured as an exclusive resource. Hypertext draws upon the paradigm of the printed page, whose navigation is an overtly individual experience; we don’t communally read. Publishers of websites provide navigation suggestions in the form of link pages, listing hyperlinks that may be of interest to the reader. Whilst there is a social element insofar as the links are recommendations from one user to another, the reader has no indication of how useful other users found the links.
In actual spaces people may navigate from maps, but often ask directions or simply follow a crowd. For example, train stations are equipped with many navigation aids such as signs and bulletin boards and yet a common way to find platforms and exits is simply to follow the crowd. Martin Svensson notes: ‘This direct and indirect interaction with other people can be thought of as social navigation, that is, in order to navigate the information space people communicate with other inhabitants of the space’ (1998: 76). Therefore, as Svensson suggests, it is possible to identify two categories of social navigation: direct social navigation and indirect social navigation.
The sheer amount of information available to us now on a constant 24-hour basis has led to a dazzling and confusing array of rumour, fact and conjecture. We therefore place great importance on filtering information. We trust editors to filter information in newspapers and on news broadcasts and we increasingly trust online journalists to filter information in the form of weblogs. We trust friends to recommend products and services and, finally, we learn by trial and error where to find good sources of information.
Whilst we are capable of automating content-based filtering, searching content for keywords and attributes, such filtering is dependent on other users accurately describing content and understanding words to have the same meaning as we attribute to them. This can produce results with varying degrees of success and accuracy. Social filtering has been proposed and implemented to combat such problems. Social filtering, such as the Amazon.com recommendation system, ‘recommends information based on what other people with similar tastes like or dislike’ (Svensson, 1998: 83). Collectively known as ‘recommender systems’, services such as Amazon’s ‘Other people who bought [this] also bought [these]’ track every user’s habits as a profile. For such systems to work a rating system must employed. Ratings can be either implicit, such as time spent reading a certain page, or explicit, a score given by the user. Thus websites do not rely on futuristic Artificial Intelligence, they are people powered.
The benefits of recommender systems are that information is filtered according to quality rather than content and many systems work simply by users navigating the website. Thus, users feel they are getting something for nothing. However, as Svensson points out, there are two dilemmas with recommender systems. Firstly, input and ratings are imperative for a recommender system to work. When the service first starts there are little or no ratings to work with and so recommendations may be poor. Secondly, it is very difficult to get users to offer unbiased ratings once it is clear that many others before them have rated an object highly. Similarly it is difficult to get users to rate a low-rated object at all once several people have rated the object poorly. (1998: 84) Recently it was revealed that many authors anonymously reviewed their own books on several of the Amazon websites (Smith, 2004), highlighting the fact that many social navigation tools rely on an element of trust, which can be broken.
In cybrid spaces recommender systems, based not only on virtual interactions but also actual interactions, may not suffer some of the problems of purely virtual systems. Trust issues may be partly answered in location-based peer-to-peer systems by the simple fact that users must be fairly close to one another. Similarly, once people are close to each other they do not have to rely simply on virtual social navigation clues. Humans are extremely adept at recognising and interpreting the continuous stream of information we each give off in the form of body language, facial gestures and many other signals. Steven Johnson equates this to ‘mind reading’ as we innately guess other people’s mental states (2001: 196). It is within such close proximity to other people that much of our cybrid social navigation will take place.
Posted by Sam at April 19, 2004 05:22 PM