For a dialogue system it seems good to have the initiative, and constrain as much as possible the range of responses that the user will give (e.g., By formulating questions such that there is a small range of possible answers), because this makes the task of interpreting the user's input easier by narrowing down the space of possibilities. This is especially useful for improving speech recognition results. On the other hand, such strict control may impede efficiency (e.g. because the system prompts get longer), and it makes the dialogue inflexible (not offering more user-friendliness than a menu-based interaction). The challenge for dialogue systems therefore is to strike a good balance of mixed initiative, possibly by adapting to the user's expertise and to the dialogue situation.
[Chu-Carroll and Brown1997] proposed two levels of initiative modelling: task vs. dialogue initiative. [Strayer and Heeman2001] propose to reconcile initiative and discourse structure.
[Chu-Carroll2000a] describes MIMIC, an adaptive mixed initiative spoken dialogue system, which employs initiative-oriented strategy adaptation to automatically adapt response generation strategies based on the cumulative effect of information dynamically extracted from user utterances during the dialogue. Second, MIMIC's dialogue management architecture decouples its initiative module from the goal and response strategy selection processes, providing a general framework for developing spoken dialogue systems with different adaptation behavior. [Chu-Carroll2000b] describe an empirical evaluation of MIMIC's adaptation strategies.
[Core
2#2
2003] describe a corpus-based evaluation of initiative in tutorial dialogue, to find that (dialogue) initiative does not correlate with students' learning gain.
Reading: [Chu-Carroll2000a];[Chu-Carroll and Brown1997]