Randomization: Randomizing form elements
To open or use this sample form, go to the Design tab, scroll down to the Your forms and datasets section, and click + then Start new form; then enable Use a sample form as your starting point and choose Randomization: Randomizing form elements from the list. You can also click here to download the spreadsheet form definition.
This sample uses two independent random draws to independently randomize the phrasing of two separate questions, and it randomizes the order of choices presented in all multiple-choice questions. The first question with randomized phrasing and choice order is designed as follows (note: if you open the spreadsheet definition yourself, you will have to scroll right to see the relevance and calculation columns):
| type | name | label | appearance | relevance | calculation |
| calculate | randomdraw1 | once(random()) | |||
| select_one yesno | likesblue1 | Do you like the color blue? | randomized | ${randomdraw1} <= 0.5 | |
| select_one yesno | likesblue2 | Do you truly love the color blue? | randomized | ${randomdraw1} > 0.5 |
This pulls a uniformly-distributed random draw from 0 to 1 into the randomdraw1 field, then uses that to control the relevance of the two "likesblue" question variations, with the first configured to show up when the draw is <= 0.5 and the second set to show when the draw is > 0.5. This gives an equal (50%) chance that each phrasing is used. Each variant also includes the "randomized" appearance, so that the choice options (Yes and No) will be presented in random order.
The second question, about the color red, uses a similar method, drawing a second random number so that the second question's phrasing is chosen independently of the first's.
A final question in the form includes a list of colors and asks which is the respondent's favorite, and a text audit field is included to capture extra meta-data about survey administration (including the exact choices and orders shown to respondents):
| type | name | label | appearance |
| select_one color | favoritecolor | What is your favorite color? | randomized(0, 1) |
| text audit | metadata | choices |
Here, the "randomized(0, 1)" appearance is used to say "randomize the choice order, but exclude the top 0 and bottom 1 choices from the randomization," so that the "Other" listed at the end of the color choices will always appear last in the list. The "choices" appearance is used to say "include choice values, labels, and order in the captured meta-data." Because meta-data is captured in this way, somebody reviewing data in the Data Explorer can press the hourglass button to layer information about question timing as well as presented multiple-choice options onto the submission detail view.
