Conclusion
Large sample size does not address selection bias; volunteer samples may systematically differ from non-volunteers in ways that affect treatment response.
View claim pageArgument
[UNDERMINE → premise #1] Volunteer participants in platform experiments self-select based on willingness to modify their social media behavior, which may correlate with lower susceptibility to algorithmic influence. The framing requires effects to generalize to the broader US adult social media user population, not just to experiment volunteers. Therefore, Large sample size does not address selection bias; volunteer samples may systematically differ from non-volunteers in ways that affect treatment response. (Warrant: Sample size increases precision but does not address systematic differences between volunteers and non-volunteers that threaten external validity.)
Premises (2)
Challenges & responses (0)
No one has tested this argument yet.
An unopposed argument is untested, not proven. Filing a rebut, undercut, or undermine is how its standing gets earned.
Pending critical questions (6)
These are challenges this argument’s reasoning pattern must still withstand. Answering them on Isonomia strengthens the argument.
- Is the literature really agreed that defects of kind K bias inferences in direction B, or is the bias direction itself contested?Open
- Does study S actually have defect D, or is the description of S inaccurate?Open
- Is the expected magnitude of the bias from D large enough to overturn S's reported effect, or is the effect robust to plausible bias corrections?Open
- Has S (or a follow-up study) performed a robustness check or sensitivity analysis that addresses defect D directly?Open
- Is this critique applied consistently — i.e., would it apply to studies on the other side of the debate that share the same defect kind K?Open
- Is H supported by independent studies that do not share defect D, such that S's defect does not undermine H itself?Open
Cited by
Cited by 1 · 1 contest
Builds on this (0)
Nothing builds on this yet.
Contests this (1)
- [DEFENSE-REBUT → k5eftq] The Allcott et al. deactivation experiment recruited participants through Facebook ads, obtaining a sample that included both heavy and light users across demographic groups. If volunteers were systematically less susceptible to algorithmic influence, the 2023 Meta experiments' null findings would be conservative estimates, not overestimates, of population effects. Therefore, Selection bias in volunteer samples would more plausibly bias toward finding effects rather than null effects, because volunteers willing to modify behavior may be more engaged users with greater exposure to algorithmic content. (Warrant: When selection bias would plausibly operate in the direction of finding effects rather than null effects, null findings from volunteer samples are if anything conservative estimates of true population effects.)contestsin a private deliberation
Embed this argument
<iframe src="https://www.isonomia.app/embed/argument/WMGniM5I" width="600" height="400" frameborder="0" style="border:1px solid #e5e7eb;border-radius:8px;" title="Isonomia Argument" loading="lazy"></iframe>
Copy and paste into any website or forum that supports HTML.
Join the deliberation on Isonomia
Support, challenge, or extend this argument with structured reasoning in Isonomia.