Interesting stuff around conversation from a human-computer interaction study. A Japanese person and an American person are placed in a virtual conversation in 3D space. A computer agent detects pauses in the conversation and suggests a new topic to proceed with. Some agents were safe, injecting topics such as movies, music and the weather. Some were unsafe, presenting money, politics and religion.

While both Japanese and American participants found the unsafe agent’s subjects more interesting, the Americans found the safe agent of value while the Japanese did not. “Of value” means increased opinion of their own behaviour and of their conversational partner.

Implications of this kind of research are interesting – they’d better be. I get to mark about 60 assignments on this stuff in a few weeks’ time.

Personally, I want interfaces to be as unobstrusive as possible and messages to be informative. More as it comes up.