Metablog

Identity and deception in the virtual community

The author examines various aspects of "identity" in virtual communities, and in particular, the role of multiple and deceptive identities. Why is "identity" so important? Because in any community, evaluating information is deeply influenced by its source, and the reputation of its members is what makes a community. The author looks at various ways that identity can be perceived in virtual worlds, despite the lack of physical clues. The particular virtual community studied is the world of USENET newsgroups.

There are two kinds of signals in any communication system. Assessment signals are those which actually require that a sender possess a trait to send a signal. E.g. a man spending money is a reliable signal of him being rich. Such signals are said to follow the handicap principle. On the other hand, for conventional signals, the sender need not possess the trait to send a signal. E.g. a man wearing a shirt saying "I am rich". While the chances of deception are obviously higher with conventional signals, they are also less costly, both for the sender and receiver.

USENET posts can reveal identity in subtle ways. The domain of the poster says something about his environment (is it from a .edu address or a aol.com address?). An institutional account is easier to trace to a real person than a commercial account.

One of the strongest hints of identity is a writer's "voice". This is a signal that is built and recognized over a long time, and regular readers can usually spot a particular poster's "style". In some cases, this "voice" is so strong that people can identify is regardless of the "From:" line of a post.

Other cues to identity can be inferred from signatures at the end of messages. Is it a geeky signature with pseudo-code snippets? Does it have a quote? An organizational disclaimer?

Yet, all these signals are easy to fake. One can fake categories (male, female, professor, student), impersonate particular individuals, or even simply omit their identity. USENET is full of trolls that pretend to be legitimate members of a group but actually want to disrupt discussions among other members. Trolls are harmful not only because they make provocative comments and give bad advice, but because they ultimately heavily dilute the level of trust in a group.

USENET and blogs are different in important ways --- blogs have a clear separation between "message" (the blog post) and "discussion" (the comments), and are generally much less chaotic and free-wheeling than newsgroups.

The most relevant part of this paper when thinking about blogs is the building of a "voice". I think the single most defining characteristic of a blog is its voice. A newsgroup is an open round-table --- a blog is a speaker at a podium. Even in group blogs, shared values among the authors result in some sort of a coherent "message". How is this voice built, and recognized? Over a period of time, just like in newsgroups.

There is one clear parallel with trolls --- comment spam. This has become an increasingly common problem over the last year. Comment spam takes the form of either outright insulting and provocative comments, or completely unrelated comments that market some product or website. Initial solutions blocked IPs of offending posters --- with dynamic IPs this was easy to get around. Newer tools use content-based comment filtering.