Hey everyone,
tldr: 1) Is there an alternative to using embedded data and loop & merge for randomly assigning text blocks to participants so that the same block can appear on several pages and still be identifiable in the analysis? 2) Is there a way to import embedded fields from excel?
I am planning a manipulation check for an experiment in which participants evaluate LLM outputs along three dimensions. In this manipulation check survey, I have 33 unique LLM outputs for each of the 3 treatment conditions to be evaluated. Each output needs to be randomly allocated to participants so that every output in every treatment is rated by three independent raters. Each rater should only see one output, but they evaluate it across three consecutive survey pages. That means the assigned output must be displayed on and remain identical across these pages and must be trackable in the data. My last attempt using loop and merge failed because the randomisation didn’t end up balanced.
My current plan is to:
- Randomly assign participants to one of the three treatments using a randomiser.
- Within each treatment, before the evaluation pages, I use a randomiser to randomly assign one of the 33 outputs as embedded data field to participants.
- Pipe the embedded data (LLM output) on each relevant survey page.
However, as far as I can tell, this requires creating 99 embedded data fields (the texts plus their identifiers) manually. I’m hoping there’s a more efficient approach.
Is there a more efficient way to implement the randomisation? Or is there a way to import embedded data fields directly, ideally from Excel? I’ve tried altering the qsf file several times, but without success.
I’d be grateful for any suggestions or ideas. Many thanks!