The Effect of Explainability on Trust
in Neural Audio Processing Systems

What is this study about?

You will listen to short audio clips processed by versions of the same AI audio compression system. In some sessions, the system shows extra details about what happened during compression. In other sessions, it only returns the processed audio. The study asks whether seeing those extra details affects trust, while also checking whether the processed audio still sounds close to the original. The whole study takes about 15-20 minutes.

Please judge each session on its own. Use only the original clip, the processed clip, and the information shown in that session. Do not compare one session with another.

How it works — 4 sessions
1
Pick an audio clip Choose a famous recording or upload your own file.
2
The system compresses it An AI shrinks the audio to a tiny fraction of its original size, then rebuilds it. Some sessions show extra process details; others do not.
3
Listen & compare Play the original and processed clips side-by-side to hear the difference.
4
Rate trust & quality Answer 12 short questions about how much you trust the current system, then give the audio a quality rating. Repeat for 4 sessions.
What you are rating
  • Trust: how willing you would be to rely on the current system to process audio accurately and consistently.
  • Audio quality: how close the processed clip sounds to the original, not whether you liked the clip itself.
Informed Consent

Your participation is entirely voluntary. Survey response data are collected anonymously and will be used solely for academic research purposes. You may withdraw at any time by closing your browser.

If you choose to enter a participant drawing or receive study credit after completing the study, you may provide an email address on the completion page. That email address is stored separately from the survey responses, used only for drawing or credit administration, and is not included in the dissertation analysis. The drawing winner will be selected in July 2026 and contacted using the email address provided.

To help prevent duplicate submissions, this site stores a completion cookie in your browser after you finish the study. The cookie contains only your anonymous completion status and completion code.

If a participant drawing or study credit is offered, only one entry is permitted per participant. Duplicate entries, repeated completion codes, or codes that cannot be matched to a completed study may be excluded.

Audio recordings: if you choose to upload your own audio, it is used solely to demonstrate the compression system during your session. It is deleted from the server immediately after processing and is not stored, retained, or accessible beyond that point. The provided sample recordings are pre-processed and contain no personal data.