[gazetriggered_2020] Should we include the GazeTriggered condition?

[adriansteffan]

The study uses 2 conditions:

• Original, which is a regular looking while listening task
• Gazetriggered, which dynamically decides on what the target and distractor are, based on where the infant looks first. The item of the first look will always become the distractor, and only THEN is the audio signal triggered, guiding the infant to the other item

For now, I have filtered out the gazetriggered trials, since it is not super clear how these would fit into our dataset because

A) The trial type is dynamically generated during the experiment based on the look of the participant (meaning it changes within the same trial as the target gets chosen). While we could assign the final trial type to the entire trial, the infant would always start at the distractor when hearing the word, according to peekbanks interpretation of the data

B) The point of disambiguation depends on the gaze of the infant and varies greatly as a result (the study has a similar issue for the original condition, but the variability is much lower) - might not be much of an issue

C) There is little mechanism to inform the consumer of the peekbank dataset of this quirk of the trials, other than the vanilla_trial field - but the exceptions listed in the column info seem like much lesser modifications to the experiment than gazetriggered

What do we do here? Ignore the condition or hack it into the peekbank format?

[mzettersten]

Yep let's filter out the gaze triggered trials completely - they just don't really fit into what we're up to with peekbank