actually has not understood; thus those strategies meant to avoid (rather than repair) misunderstandings may be useless in HCI. In contrast, direct explicit accounts, such as explanations of the system's malfunctions, are more straightforwardly employed and only presuppose the recognition of critical situations. For instance, comparable to speakers' accounts of face-threatening acts in conversation, if the system detects changes in speaker behaviour which may be caused by a changing attitude towards the system, it may calm down the user by explaining its shortcomings or by apologizing for them. It may also be very effective to sum up the current state of the discussion, but this again requires elaborate capabilities including dialogue memory and the ability to compare the current state to a projected goal. Consequently, considering the difference between HHC and HCI, the least costly and most efficient way to influence the speakers's attitude towards the system, which also has been proven to be effective in the Wizard-of-Oz experiments, may be to use explicit accounts which can be previously generated and which only require that the system recognizes situations which make such accounts necessary.
Three ways of addressing the problems caused by speakers' reactions to system malfunction which are manifest in repetitions, reforinulations, and emotional reactions have been presented: Automatic speech processing systems can be adapted to the peculiarities of utterances of this type, speakers can be explicitely instructed, and they can be subtly guided by means of features of natural conversation. For the latter alternative it was examined in how far strategies from HHC can be useful for HCI design, how costly an implementation would be and what could be gained. It can be concluded that while it is often impractical to instruct speakers before they begin their interaction with an automatic speech processing system, both alternatives, the adaptation of the system to real conversational conditions such as reformulations and emotional involvement and the employment of some of the less costly strategies speakers use in HHC, should both be followed.
Brown, P. & Levinson, S. ( 1987). Politeness. Some Universals in Language Usage. 2nd edition, Cambridge: Cambridge University Press.
Fischer, K. ( 1998). A Cognitive Lexical Pragmatic Approach to the Functional Polysemy of Discourse Particles. PhD Thesis, University of Bielefeld.