issues with experimentation (CHECKLIST)

[BenedictWilkins]

Various things need updating to make experiments easier to analysis later.

• give an option to place eye tracking events in a different file (they flood the log file)
• standardize the input event format
• when logging changes, also log the current value (e.g. fuel change -1, value 100) this makes it easier to process the file and is less error prone.
• have a mode where agents give "fake" feedback (not shown on the interface, but logged) so that we can do counter factual analysis more easily.
• show more obvious warning if the eye tracking system crashes so that the experimenter can be aware!
• position of the target in the tracking task is not in window coordinates. The position of the tracking task should be given in a log file. Complete window information (can use window_properties in init.py) should be logged.
• when eyes move out of the ICU interface window this should be known, add a new label "outofbounds"
• eye tracking timestamp is wrong, it is current "time since eye-tracker has been on"
• it would be good to include a start up or "ready" event with a timestamp
• record all mouse and keyboard input regardless of whether it is a "valid" interaction.
• there should be an event triggered when a task is "in failure", this is only true for the fuel task, make it so for the others.
• the "cause" property of an event is difficult to parse, it should just be the ID of the causing event.