Just got a Nokia N80, thanks to my friends at Nokia Research Palo Alto. I may write a longer review of it later — I like it a lot, and will like it even better when Zonetags catches up to it.
But here I want to comment on one odd design choice.
My recent Nokias have had voice dialling, which was erratic. The caller trains the phone with voice tags for selected contacts. I then had to remember exactly how to say a name the same way every time for the phone to recognize it, and differences in environmental acoustics affected it substantially.
The N80 is different. Instead of the user training it for specific numbers, it uses voice recognition to try to match any effort with the phone book. However, there is no way for the user to give it feedback and help it improve its recognition. Trying it out, I found that when I named a friend I call all the time, it wanted to call the optometrist I call once a year, and persisted in its error. I finally had to change the entry for my friend so that the names (which are not particularly close) seemed less similar to the N80.
So the user has no opportunity to give the phone feedback or to override it (other than to cancel the call that it tries to make). In my teaching I often include the topic of configuring the user: what the design assumes about the user, and which tasks are delegated to the user or arrogated to the technology. In this case, the technology claims ownership over matching the user’s voice with the contacts list. This is particularly frustrating since I need to use the voice commands when I’m least able to fiddle with the phone, such as while driving. In addition, the computer-generated voice with which the phone announces who it’s about to call sounds like a seriously hearing-impaired person trying to speak, but only marginally succeeding. This does not inspire confidence — I feel like I’m delegating a task to a being too impaired to fulfill it, with no ability to learn.