Skip to content

InterView

InterView is a platform for conducting survey interviews in virtual reality. An interviewee wears a VR headset and sits across from an avatar in a shared virtual room; the interviewer sits at an ordinary web browser. The two are connected by a live audio/video link, and everything that happens - speech, gaze, posture, hand movement, answers given - is recorded in a single, time-aligned data set.

It was built at the Text Technology Lab for the FACES project (Feasibility, acceptance, and data quality of new multimodal surveys, DFG project 539621548), part of the DFG priority programme SPP 2431 New Data Spaces.

Why it exists

Survey methodology has a measurement problem it cannot see. A telephone or web survey records the answer and nothing else - not the hesitation before it, not where the respondent was looking, not whether they understood the question. The interaction that produced the answer is discarded, and with it most of the evidence about whether the answer is any good.

InterView keeps that context. Because the interview happens inside a simulation, every channel is observable at once and on a shared clock:

  • what was said, by whom, and when
  • where the respondent was looking, continuously
  • head, hand and body posture throughout
  • facial expression, via the headset's face tracking
  • the answers themselves, and how long each one took

The research question is whether a VR-mediated interview is feasible, whether respondents accept it, and whether the resulting data quality holds up against conventional modes. The platform is the instrument that makes asking that question possible.

What has been built

  • VR client

    A Unity application for Meta Quest headsets. Renders the interview room and the interviewer's avatar, streams the participant's microphone, drives the embedded questionnaire, and logs every tracked channel. Built on Va.Si.Li-Lab.

  • Web client

    Two browser pages - one for the interviewer, one as a non-VR fallback for the participant. Camera and microphone preview, device selection, virtual background, live chat, gaze overlay, answer options, and recording control.

  • Media layer

    A Janus SFU carrying audio and video between headset and browser, plus a data channel for control messages. Optional self-hosted TURN relay for difficult networks.

  • Data layer

    A room/session server keeping VR state in sync, and a logging API that writes every sample and event into MongoDB for later analysis.

How a session runs

  1. A participant is issued a token - the identifier for their interview.
  2. The token is entered in the VR client. It deterministically derives the media room, the VR room, and the questionnaire URL, so headset and browser meet in the same place without any coordination step. See Token and room derivation.
  3. The interviewer opens the same token in a browser and joins.
  4. The interview is conducted. Audio and video flow through Janus; the interviewer can push answer options, start and stop recording, and follow the participant's gaze live.
  5. Recordings land on the media server; all tracking and event data lands in MongoDB, keyed by the same token.

Where to go next

I want to… Go to
Understand how the pieces fit together Architecture
Deploy the whole stack from scratch Setup Guide
Configure the media server Janus
Set up a TURN relay coturn
Build and configure the VR client VR Client
Retrieve recordings, or debug a session Operations

Two documentation sites

Documentation for the VR client itself - Unity setup, scenes, avatars, tracking, the Va.Si.Li-Lab framework InterView is built on - lives in the Va.Si.Li-Lab documentation. Everything about the InterView platform and its services is here.

Citation

InterView builds on Va.Si.Li-Lab. If you use it, please cite:

@inproceedings{Mehler:et:al:2023:a,
  author    = {Mehler, Alexander and Bagci, Mevl{\"u}t and Henlein, Alexander
               and Abrami, Giuseppe and Spiekermann, Christian and Schrottenbacher, Patrick
               and Konca, Maxim and L{\"u}cking, Andy and Engel, Juliane and Quintino, Marc
               and Schreiber, Jakob and Saukel, Kevin and Zlatkin-Troitschanskaia, Olga},
  title     = {A Multimodal Data Model for Simulation-Based Learning with Va.Si.Li-Lab},
  booktitle = {Digital Human Modeling and Applications in Health, Safety, Ergonomics and Risk Management},
  publisher = {Springer Nature Switzerland},
  address   = {Cham},
  pages     = {539--565},
  year      = {2023},
  doi       = {10.1007/978-3-031-35741-1_39}
}