I believe that this post didn’t receive nearly as much attention as it should have.
I know that it might be tricky to make this happen, but it would be really useful to have a fail-safe in online experiments that would allow a researcher not to lose participant’s data if he or she accidentally closes the window/tab before the experiment ends naturally (either by pressing Esc or some other key). I am writing this because I just lost a very useful chunk of data purely accidentally.
Now, @jon mentioned in one of his posts that “we have to assume that when a participant quits the browser they do mean to be excluded from the study and we should not be keeping their data.”. However, there are plenty of scenarios when this is not the case. Many experiments are developed in such a way that they end with a farewell message after solving the necessary tasks, giving the participant a feeling that he is done and that he can close the window now. There might be a message that he should press a key or something but this can be easily overlooked, especially after a cognitively demanding experiment. Closing the window, in this case, would probably not mean that the participant does not want his data to be stored.
One example is my experiment, the nature of which simply does not allow me to predict the exact moment when it should be ended. Furthermore, it is quite cognitively demanding. Before the (possible) end of the experiment, there is a long interview, and only after the interview, it is clear whether the experiment is to be continued or not. Consequently, there is a high risk for me to lose the participant’s data since I can simply forget to warn the participant once again, and the participant can simply forget the previous warning.
Data loss is a serious problem, and there should definitely exist a way to restore it in the cases of accidents. For example:
If privacy is of main concern, then there could at least be an option to return back to the point from where you left off the experiment before accidentally closing the window/tab. Perhaps a unique hyperlink can be created for such purposes. In this case, a participant could be warned and he could return to the experiment and make sure to end it properly.
Or, if this is technically too ambitious, data could nevertheless be stored for a while (e.g. 2 hours) when a participant doesn’t finish the experiment properly, and there should be a possibility of retrieving this data by making some sort of request from the side of the researcher as well as the participant.
These would probably not be viable solutions for experiments that accept anonymous participants, but, for those that make us of participants whose “identities” are known (in one form or another), they probably would.
Hopefully, someone will look into it. I am sure you have thought about it already, but … is there really nothing that can be done?