Prompt generation guidelines

Example prompt:

With respect to frequency of contractions, characterize the difference between antibiotics and placebo.

Task description:

This task involves generating questions (or "prompts", see example above) that are answerable on the basis of a given full-text article describing a randomized control trial (RCT). More specifically, such prompts will concern the comparison of specific interventions with respect to particular outcomes. Essentially, these should ask whether a given treatment was found (in the RCT) to be significantly more effective than a comparator treatment with respect to a defined outcome of interest. Trials typically report results for multiple outcomes (and often more than two interventions), and the results for these will of course differ. Therefore, you will be asked to generate multiple unique prompts for each full-text article. Taken together, the answers to these prompts should summarize the main findings reported in a given article regarding the treatment effectiveness of the interventions studied. A high-level description of the task is as follows.

For each article:
  1. Find an outcome, comparator, and intervention.
  2. Add these, along with the relationship between each, to the table.
  3. If there are no more "prompts," then submit. Otherwise, repeat the process.