Treatment response — small points, wide intervals

Across the represented continuation-probability treatment, fixed-panel point differences are +0.083 (S2-absent) and +0.078 (S2-present). Conservative exact simultaneous 95% intervals are [−0.171, +0.330] and [−0.181, +0.330].

The treatment changed both the continuation process and the text communicating it — round-one actions identify response under a specified representation, not a semantically neutral economic parameter. The data establish neither equivalence nor a null nor a narrow bound. A design-effect heuristic (six episodes per prompt, between-share 0.855–0.961) gives roughly 16.5–18.2 episode-equivalents per condition: prompt-family size, not raw episode count, is the operative precision constraint.