We conducted a first user study with 7 inhabitants and 2 caregivers of 'Haus Bersaba'. Since we are in the prototyping phase of our project where we have no fully functioning system, yet, we use the Wizard-of-Oz methodology: The user gets the task to brush his/her teeth at the washstand equipped with two cameras, a microphone and a TFT display with speakers. He is not aware of being supported by a caregiver but thinks that he is faced with an automatic prompting system. However, the caregiver - the wizard in our scenario - operates the system via a graphical user interface (see image on the left) and can't observe the washstand directly. Instead, he gets the live-streamed images and audio from the sensors installed at the washstand. The caregiver assists the user in the brushing task by generating either audio or combined audio/video prompts via the GUI. The prompts are delivered to the user via the TFT display and speakers in realtime.
We conducted six WOz trials in total with three persons performing two trials each. We are interested in the user's reaction behavior to system prompts in comparison to prompts given by the caregiver directly. Hence, we also conducted five trials performed by five persons where the caregiver directly prompts the user. We will refer to the two scenarios with WIZ (for system prompting generated by the wizard) and CG (for direct caregiver prompting), respectively. We found that the average reaction behavior of the users were comparable in both scenarios. However, the average results hide significant variations amongst user, especially in their particular reaction behaviors to prompts of different modalities: One user was highly distracted by combined audio/video prompts whereas another user was able to react correctly to audio/video prompts in all cases. Detailed results can be found in the paper Prototyping of an automatic prompting system for a residential home