Fig. 3
From: A meta-analysis of the persuasive power of large language models
Meta-regression estimates for moderator analyses. Each plot shows the estimated difference in effect size (g) with \(95\%\) (thin light blue bars) and \(90\%\) (thick dark blue bars) CIs for (a) LLM model, (b) conversation design, and (c) domain. The numbers in parentheses indicate the number of effect sizes (n) contributing to each estimate. Each moderator comparison is based on small subgroup sizes (see n values in parentheses); results should thus be interpreted cautiously.
