The following suggestions are based on my experiences, and the feedback received from thousands of Workers.
When possible, design the task within the HIT window.
- If this is not possible, then make sure you choose external websites that supported by multiple operating systems and browsers (such as Qualtrics or Survey Monkey).
- Add target="_blank" to the HTML code for these external links so they open a new window to minimize disruption to the Worker.
Make HITs fun and interesting.
- Use games, puzzles, and stories to get the information you want.
- Include images and visuals whenever possible (e.g., on the "Thank You" page, periodically throughout a survey).
Be clear with your instructions.
- Show all of the steps required to complete the HIT. Use numbered lists and visual aids. Ask someone else to read them in the pilot test. For more complicated HITs, include a short video to demonstrate how to do the HIT.
- Specify how much time it takes to complete the HIT, based on the average time from your pilot test (see below).
- Be clear about whether Workers can complete more than one HIT in a batch.
- Explain under what circumstances a HIT will be rejected. Tell Workers the time frame for approving/rejecting submissions (e.g., 24 hours, 3 days, 1 week).
- Tell Workers who to contact if they have problems.
Design HITs to be "user friendly".
- The format should be simple. Use font that is size 12 or larger, and avoid unusual color combinations (e.g., yellow font on a black background).
- Allow Workers to check their eligibility before accepting a HIT. Use a Worker ID Check or provide a list of ineligible Worker IDs.
- Include a progress bar so Workers know how to pace themselves.
- Limit scrolling and limit the total number of pages. Avoid putting only one item on each page, and avoid putting all of your items on one page. Balance the use of "bubbles" (multiple choice questions) and open-ended questions.
- When asking about potentially sensitive information (e.g., gender, race, religion, etc.), provide response choices like "Other" or "Prefer not to answer".
- Be careful with attention checks. Avoid quizzing Workers as a way of checking whether they read something (unless you told them to remember it).
- Specify whether a task will require Workers to listen to audio and video clips. Some individuals are deaf, and could benefit from this information up front.
If you are using an external website (e.g., Survey Monkey) to collect information from Workers, the last page of the survey should include a debriefing page to explain what the HIT was about, contact information, and completion codes.