Individual-level randomization does not systematically underestimate population-level effects when treatment and control groups share network connections, as spillovers would contaminate the control group and bias results toward null.

[DEFENSE-UNDERMINE → vefpss premise #2] The Guess et al. (2023) experiment included users embedded in real social networks where treated users' changed exposure could influence untreated network contacts. If network spillovers were substantial, control group members would be partially treated through their connections to treated users, biasing the experiment toward finding null effects. The null result therefore either reflects true absence of effect or underestimates the true null due to spillover contamination, not underestimation of a positive effect. Therefore, Individual-level randomization does not systematically underestimate population-level effects when treatment and control groups share network connections, as spillovers would contaminate the control group and bias results toward null. (Warrant: Network spillover effects in randomized experiments bias toward null findings by contaminating the control group, so invoking spillovers cannot rescue a positive causal claim from null experimental results.)

• The Guess et al. (2023) experiment included users embedded in real social networks where treated users' changed exposure could influence untreated network contacts.
• If network spillovers were substantial, control group members would be partially treated through their connections to treated users, biasing the experiment toward finding null effects.
• The null result therefore either reflects true absence of effect or underestimates the true null due to spillover contamination, not underestimation of a positive effect.

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

No one has cited this argument yet.

No arguments cite this one yet — no one has built on or contested it. That is an absence of engagement, not a finding of soundness. Build on or contest it on Isonomia to change that.