The Internet Survey Lab at the Center for Bioethics and Social Sciences in Medicine (CBSSM), led by Dr. Brian Zikmund-Fisher, facilitates the programming of complex experimental designs, using the graphical and interactive capabilities of the Internet. CBSSM has extensive experience in developing, programming and conducting survey research using Internet-based methodologies.
Why We Use the Internet
A key advantage of Internet surveys is that they can shape and direct a user's experience in response to computer generated randomization and/or respondents' own answers to questions earlier in the survey. Additionally, page and answer order can be truly randomized as appropriate to limit cognitive biases. The unique advantage of Internet surveys, however, is that many different types of stimuli can be randomized or varied; static visual images, movies, or sounds can all be used in addition to text. Furthermore, the nature of the browser interface enables user-directed interactivity, such as user-adjustable risk communication graphics, that provide unique opportunities for both knowledge communication and response assessment.
Using the Internet to conduct survey research is also very efficient: we can develop and test surveys in only a few months' time, and once a survey is ready, large scale data collection (e.g., 1500-3000 completed surveys) can be completed in only 2-3 weeks. Such surveys can also be cost effective, since while significant effort goes into development, creation, and testing of the survey, almost no personnel effort is required for data collection, entering, coding, or cleaning. In addition, oftentimes several small surveys can be combined into a single instrument, creating further efficiencies.
Sometimes, our studies use large, demographically diverse samples obtained through commercial survey research firms. This methodology allows us to tailor the population being surveyed on multiple demographic variables (e.g., sampling only women age 40-75 for a study about breast cancer treatments) and provides us with ample statistical power to conduct multi-factorial experimental tests. Other times, we use more inexpensive samples from Amazon Mechanical Turk (MTurk) for quick pilot testing or to enable rapid, iterative testing of designs. Regardless, the use of randomized designs ensures high internal validity for the research despite the use of an Internet-only sample.
CBSSM has had considerable success using this methodology, publishing multiple manuscripts in highly regarded peer-reviewed journals. Studies that have used this methodology have addressed a variety of topics, including:
- The use of pictographs to display risk (2008, 2008, 2014) including in comparison to other graphical formats (2008, 2010, 2010).
Note: to create your own pictographs, see www.iconarray.com.
- Misprediction of happiness between younger and older adults (2005)
- Elicitation of utility and willingness to pay (2007, 2007, 2008)
- Research ethics, e.g., participation of mentally vs. medically ill in research (2005)
- Risk communications that emphasize incremental risks instead of absolute risks (2008)
- Simplifying risk communications about adjuvant therapy options (2008).
- Effect of risk labels on prenatal screening decisions (2007).
- Time-insensitivity in people's understanding of survival curves (2005, 2007)
- Self-other discrepancies in medical decisions (2006, 2008)
- Sequential vs. all at once presentations of risk information (2011)
- Testing of animated or interactive risk graphics (2011, 2012, 2014)
- Optimal levels of precision in risk communications (2011, 2012)
- Framing of health promotion messages (2012)
- Exploration of role of narratives in decision making (2010)
- Values Clarification (2015)
- Intuition and Deliberation in Decision Making (2015)
For questions about our methods or inquiries about potential Internet survey research collaborations, please contact Brian Zikmund-Fisher at firstname.lastname@example.org.