Prevention and reduction of mistrust and reactance in user-companion-interaction processes
In the first funding period Wizard-of-Oz experiments with realistic dialogues were conducted and supplemented by user interviews within the carefully designed LAST MINUTE scenario. Thereby the LAST MINUTE corpus (N = 130) was created, comprising multimodal recordings (audio, video, psychobiological data) of the experiments, verbatim transcripts (in GAT 2 standard) as well as data from psychological questionnaires and interview recordings (n = 73).
Analyses of subject utterances in terms of politeness, vocabulary, verbosity, language preferences and paralanguage together with their temporal change confirm a great variety of approaches to the system and indicate group classification: e.g. elder subjects were significantly more polite but also less successful than the younger.
On the other side, qualitative analyses of the interviews give insights about subjects’ experience of the interaction. Cooperativity has thereby emerged as essential for the success of a user-companion interaction. Particularly, reduced cooperativity and defense reactions and even mistrust might occur, when the user ascribes negative intentions and attitudes – e.g. by system-sided pressure, coercion or subjugation demands – towards the system.
The planned empirical studies will, by systematically varying the LAST MINUTE scenario, investigate the effects of trust-promoting measures (e.g. anthropomorphization) in relation to specific user groups, and how to ensure that no “Uncanny Valley” effects occur within the Companion-technology.