Error "The experiment failed to load"

KirstenStark

Hi,

sorry for the confusion, I just switched here from a GitHub issue where I was describing a problem in which, since the last Jatos update, a significant amount of participants, in the middle of the experiment, suddenly sees a blank screen with the info "The experiment failed to load". One participant sent me screenshots of her console, and it seems like images couldn't be loaded (across participants, the experiment doesn't always stop working on the same image or position in the experiment (not even the same jsPsych component), and all images are loaded perfectly for some participants).

Kristian suggested that something was broken, either the JATOS server or something between the server and the participant (router, network etc.).

So I contacted our admin. He went through the server settings and told me he couldn't see anything wrongly configured or any evident problem that might be causing any error. Nor is there an error message in the study logs.

I find it difficult to believe that it is a problem on the participants' side because it now happens in more than 25 % of our participants, and we never had this particular problem before the update. We also did not change anything to our current study. Furthermore, emptying the cache doesn't help with this problem, and it happens in different browsers and on different OS. So I assume it is probably an interaction between the server and our studies?

Is there anything that might have changed in the communication between study and server since the last update? Or is there any way I could test were the problem emerges from?

Several things I was thinking about:

1) All images are preloaded in the cache and there are no preloading failures logged in the participants experiencing problems, so normally the images shouldn't be lost that often (?). I also think that previously, in case an image wasn't loaded properly, the trial was skipped, and the experiment continued to run (there is a timeout for each trial, so the experiment should just proceed to the next trials). Is there a possibility on the JATOS

side to force the experiment to continue in case of loading failures?

2) Worst case, for now, if one component stops working, I would even be fine if the experiment just jumps to the next component and continues from there. Is there maybe a JATOS code for that?

Thank you so much in advance!

Best

Kirsten

elisa

Hi Kirsten,

It might take Kristian a couple of days to reply, and I know you're in the middle of running your study, so just a couple of points asking for more information for now:

You said there are no error messages in the log. Could you send us the log at around the time your participant had the error?

Do you have any other hints, any idea of how you could reproduce the problem? (actually, given that we both have access to the same server, could you make a clone of your study and make me a member of that cloned version? I can try to see if I notice anything)

There is a code to jump to the next component.  jatos.startNextComponent() will not send any data but allow your participants to continue. Jumping to the next trial and error handling might need to be done from within jsPsych. But I agree that it sounds like it might be an error related to JATOS, maybe a combination of JATOS and jsPsych.

If you give me access to that copy, I'll try to see if I get any more clues.

Best

Elisa

KirstenStark

Hi Elisa,

thank you!

I am attaching the study log of one of the participants who had the problem I described. To localise any technical issues we might have, we started to upload the data after every trial, which is the reason why the log is that long. For comparison, I'm also sending a log of a session that worked perfectly fine (resultID 23684).

I don't really have any idea how I could reproduce the problem. One of my students had the problem twice, the third time, after she emptied her cache, the experiment worked perfectly. However, this week we had 2 participants out of 3 who ensured us that they had emptied their cache and still had the same problem.

I created the clone, I just need your username to add you as a member. I will email you for that.

Best

Kirsten

complete_study_log_workerID23684.pdf

study_log_experiment-that-stopped-working_workerID23681 Kopie.pdf