Pro-social teasing and cmcBerkeley psychologist Dacher Keltner covers some of the social benefits of teasing in the NYTimes Magazine today: indirect communication of group norms (such as the frat boys who use a nickname to hint that you shouldn't drink *too* much), collective stress release through laughing (even for the target), and flirtation. He frames it in Brown & Levinson's politeness theory, especially the "off record" part of their framework, where you never explicitly say what you mean, you hint.
So what about teasing in computer-mediated communication? Will we ever develop the skills to signal and interpret playful teasing online, or are we stuck in binary world of vanilla literalism and cyberbullying? I don't think that's true, but we already know that people are pretty lousy at interpreting tone in CMC, and there was certainly an element of mixed signals in the Megan Meier case. (Had Megan met "Josh" in person, her feelings might have been less intense than those constructed from text only, and had there been more face-to-face contact between Megan and Lori Drew's daughter, some of the deference signals Keltner mentions (such as blushing and head lowering) might have convinced the Drews to back off. But that's just morbid Monday morning quarterbacking, and this is a pretty extreme case.)
But on to happier topics: Can we detect friendly joshing and flirting online? Sure (to an extent), but what about from people we've never met face-to-face? Do you have to be able to picture the speaker smiling while you read the text or imagine his voice? And what are the mechanisms for social norming (like with the frat house morality tale) online? Do you just monitor your facebook feed to see what's acceptable in your friend circle? Do we have to pepper everything with emoticons? Can we build machine learning tools to detect it automatically, say, so parents or Facebook (barring privacy issues right now) can distinguish between the good kind of teasing and potential harassment? The few politeness detectors out there don't include any of B&L's off-record markers (the on-record markers, like inclusive language or token agreement are hard enough). I'm looking forward to the next decade of communication research as our online behavior norms start to cement and as our tools for analysis get better.