I'm looking for documentation on how people react to talking to an artificial intelligence system like, say, Siri (especially now that iOS6 has dictation). How 'comfortable' is it to use voice commands? Is there a psychological barrier to it?
The documented barriers I found deal mainly with an expected error rate of 1 in 3 words that the agent misinterpreted, and the need to manually correct those errors in order to move on. Although speech input is 3x faster than typing, this error rate leads to frustration and eventual abandonment of speech input.
Source: Kumar, Paek and Lee's "Voice Typing: a new speech interaction model for dictation on touchscreen devices", CHI 2012.
Your question is quite interesting. There are some communication barriers while the user is using such AI systems.
When any user gives any command to the AI speech assistant (like Siri in iOS) then it may happen that the user may speak in a different accent leading to confusion as that accent may not be / or misunderstood by the speech assistant.
Hence we can say that there are some psychological barriers to it.
Obviously it is easy and comfortable to use the speech assistant but it may leads to frustration if the user commands are misinterpreted or misunderstood by the speech assistant.