Hi all,
Background:
I have a sparse sampling video task - meaning that a video is paused at semi-random intervals to ask questions - programmed in jsPsych (v7.3.3 with some modifications). The stimuli are sixteen 2m47s .mp4 videos compressed and optimized for online streaming via Premiere Pro. File sizes range from 24.1MB to 49.9 MB. The videos, when played together make up a 44m murder mystery that I have my participants watch and the segments play in sequence. This was necessary for loading and storage purposes. I determine when the video will pause by ascribing each subject with a unique ID that increments with each successful subject (via the @wakecarter vespr tool and some hardcoded timestamps associated with possible IDs). To spread out loading times, I preload one video at a time and if a video transition (i.e., switching from one 2m47s to the next) is scheduled to occur before the next pause, I preload that video while subjects are answering questions. I have this project stored on Gitlab, running through Pavlovia, and subjects are recruited via Prolific. It also only runs via Chrome, Firefox or Safari because those were the browsers I could reliably test.
Issue:
When operating this experiment locally, in pilot, and formally running, I encountered no issues and was able to make it through the entire task. However, when I opened recruitment up on Prolific, several subjects reported that video loading timed out and they were not able to complete the experiment. There did not appear to be any consistency in the browsers or operating systems, but admittedly many subjects ghosted me before I could get more details. Their data was not retained on Pavlovia and although I output console logs as part of the experiment, none of the participants were able to share console logs with me. Eithout logs or data, I can’t be sure which study settings they experienced the issue under. Fundamentally, my issue is I can’t figure out why I’d experience an error, I can’t replicate the error, but subjects keep reporting the error.
I’ve tried multiple computers, operating systems, browsers, etc. I throtled the number of subjects who could access the task simultaneously. I broke the video into the smallest reasonable size I could and I still have subjects reporting the issue at random points. If I didn’t have so many people (9 so far) experiencing this issue independent of one another, I’d be inclined to blame user error or nefarious intentions.
I’m wondering if anyone has any better ideas for more robust video loading/streaming or to get this error to replicate for me. Of course, if I’d forgotten any information that might be helpful to know, please let me know.