Feasibility, acceptance, and data quality of new multimodal surveys
Aim
The FACES project (Feasibility, Acceptance, and Data Quality of New Multimodal Surveys) aims to create a multimodal data space for survey research that can expand and replace face-to-face interviews in the future through the use of virtual reality (VR) and artificial intelligence (AI). This multi-interface system for online surveys is designed to offer a high degree of variability in terms of avatars, situational parameters, interfaces and AI technologies for the automatic processing of speech and behavioural data.
Background
Interview-based surveys face challenges such as increasing costs and decreasing response rates. Recent innovations in the fields of VR and AI offer new approaches to address these issues. Avatar-based interviews open up additional degrees of freedom through the choice of avatars and the variability of interaction situations. However, systematic studies on the effects and acceptance of such systems and on their potential to reduce interviewer effects are still lacking. The project goes beyond existing studies by comprehensively investigating self-avatar and other-avatar effects and their influence on classical interviews for the first time. In addition, scenarios with different degrees of immersion (from fully immersive VR interviews to video-based interviews) are compared.
Approach
In a first step, an open-source system for avatar-based and video-based interviews is being developed. This system will be used in a small preliminary study to examine the effects of different avatar and situational characteristics in experiments. Scenarios with different degrees of immersion (from fully immersive VR interviews to video-based interviews) will also be compared.
Based on the results, promising feature combinations will be tested in interviews with former NEPS participants.
Three central questions will be investigated:
- What are the advantages of avatar-based interviews compared to video-based interviews in terms of acceptance, feasibility and data quality?
- Which combinations of features reduce interviewer effects and how do they interact?
- How can the results be integrated into a theory for training virtual interviewers?