Note: Summarized by Claude 3.5 Sonnet based on my notes from reading the paper. For more details, please refer to the original paper. Highly recommended.
Paper: Aridor, G., Jiménez Durán, R., Levy, R., & Song, L. (2024). Experiments on Social Media. https://doi.org/10.2139/ssrn.4893781 .
1. Sample and Recruitment
- Consider using social media platform’s targeting capabilities for recruitment
- Evaluate the need for quota sampling or post-stratification
- Be cautious about claiming full representativeness of the sample
- Implement measures to prevent or detect link sharing
2. Intervention Design
- Choose appropriate intervention type:
- My favorites are…
- Encourage change in on-platform behavior
- Manipulate participants’ experience through platform features
- Provide exposure to content off-platform (but consider trade-offs between control and ecological validity)
- My favorites are…
- If using advertisements, account for potential low attention from users
3. Data Collection
- Evaluate options for data collection (e.g., browser extensions, data donation)
- Consider scalability and potential experimenter demand effects of chosen method
- Plan for longitudinal data collection if possible
4. Statistical Power and Effect Sizes
- Anticipate potentially small effect sizes
- Plan for large sample sizes or exploit within-participant variation
- Consider using repeated-measure designs to increase power
5. Attrition and Compliance
- Develop strategies to minimize attrition
- Plan for potential non-compliance (e.g., reactivation in deactivation studies)
- Consider incentivizing compliance, but be aware of potential effects on results
6. SUTVA (Stable Unit Treatment Value Assumption) Violations
- Assess potential for interference between units
- If relevant, use cluster-randomized design or two-stage randomization
- Plan to measure or account for spillover effects
7. Platform Dynamics
- Consider potential unobservable actions by the platform
- Account for algorithmic responses to interventions
- Plan for potential equilibrium and long-run responses
8. Interpretation of Results
- Be cautious about generalizing results (partial vs. general equilibrium effects)
- Consider complementing experimental evidence with observational studies
- Plan for measuring long-term effects (e.g., using surrogate indices)
9. Ethical Considerations
- Ensure proper informed consent procedures
- Address privacy concerns in data collection and storage
- Consider potential negative impacts on participants’ well-being
10. Replicability and Transparency
- Document all aspects of the experimental design and analysis
- Consider pre-registration of the study
- Plan for sharing data and code (within ethical and legal constraints)
Remember to review and adjust this checklist based on the specific context and requirements of your social media experiment.