I think the best way for me to explain my problem is by first giving a quick overview of the experiment. The task has three parts;
- In part 1, participants are presented with 2 scenarios (randomly sampled from a pool of 20 items) and write down two very common behaviours that someone may perform in each given context
- In part 2, participants are randomly presented with another 2 scenarios (sampled from the same pool, minus the first two options they sampled in the last round) and write down two very uncommon behaviours that someone may perform in that context
- In the final section, participants will have their responses fed back to them and make a series of judgments about the behaviours they listed for each scenario
^Important note: these three sections are intentionally being kept separate. It would confound the experiment if I had participants listing uncommon and common ideas at the same time, or piped their answers back to them immediately after each scenario to make their judgments.
Therefore, here are my two problems:
- Linking the pools in parts 1 & 2 so that the scenarios sampled in the first task cannot be sampled in the second one
- In part 3, how to ensure that participants are only presented with the blocks that pipe back the questions & behaviours they answered in 1 & 2, and not any of the other scenarios.
To address the first challenge, I tried using two levels of randomisers. Specifically, I used a randomiser to present a scenario from the pool of 20 (without replacement), and then another randomiser to present either the common or uncommon version. Pros: prevents the same scenario being sampled for both the common and uncommon task. Cons: (1) merges the common & uncommon task when I need to keep them separate (& have instructions to present in between task 1 & 2), & (2) doesn't guarantee participants will be given 2 common questions and 2 uncommon questions to answer overall.
I haven't yet found a method to trial for solving the second challenge.