We walk through several properties that are typical for complex systems and specifically for social media and how individual-level approaches to infer causality may be misguided and how that can be mitigated:
Comments
Log in with your Bluesky account to leave a comment
And even if the methods are well suited for the outcome measure in question, we argue that the treatment should involve an actionable interventions, such as algorithmic changes or alterations of the platform affordances instead of broad interventions that are unrealistic to implement.
Finally, we suggest to look for inspiration in fields like public health or climate science to develop a triangulating between experimental methods on the individual level, observational methods and causal inference on the collective level, and theory and computational methods to bridge the gap:
Comments