Studies of human persuadability and malleability show substantial unpredictability and chaos in human responses.

Judge as of 2026-10-07 against the research on persuasion and attitude change (meta-analyses, large experiments, reviews). TRUE: that research generally finds that how individuals respond to persuasion is highly variable and hard to predict, e.g. effects are small and vary a lot across people, messages and settings, and models predict individual responses poorly. MOSTLY TRUE: the research largely shows this, but with notable exceptions where responses are fairly predictable. MIXED: the evidence is roughly split, with some lines of research showing unpredictability and others showing reliable, predictable persuasion effects. MOSTLY FALSE: the research mostly finds persuasion effects to be consistent and predictable, and only narrow findings support 'unpredictability and chaos'; or the claim states as established something the research leaves inconclusive. FALSE: the research clearly shows that human responses to persuasion are reliably predictable.

Set before the research began.

1. E1 Message Design Choices Don't Make Much Difference to Persuasiveness and Can't Be Counted On—Not Even When Moderating Conditions Are Specified Frontiers in Psychology (per supplementary link on the fetched page; publisher not otherwise shown) · 29 June 2021

   “Persuasion message variations have small and highly variable effects.”

2. E2 How Experiments Help Campaigns Persuade Voters: Evidence from a Large Archive of Campaigns' Own Experiments American Political Science Review, Cambridge Core · 4 November 2024

   “Assessing these predictions of when ads will be more or less effective across 617 real advertisements, we find at best limited evidence for any of them.”

3. E3 A meta-analysis of the persuasive power of large language models Not shown on fetched text (PMC-hosted peer-reviewed article) · 12 December 2025

   “Although based on a small number of studies, this suggests that differences in LLM model, conversation design, and domain are important contextual factors in shaping persuasive performance, and that single-factor tests may understate their influence.”

4. E4 [PDF] Effective strategies for changing public opinion: A literature review Sentience Institute · 8 November 2021

   “Public opinion surveys usually find that some demographic variables are statistically significant predictors.”

E1, a review of persuasion research, finds that message variations have "small and highly variable effects". E2's study of 617 real campaign ads finds "at best limited evidence" for theories of which ads will work better, so persuasiveness is hard to predict. E3 reports that contextual factors such as model, conversation design and domain shape how persuasive an LLM is. E4 notes that public opinion surveys usually find some demographic variables to be statistically significant predictors, so opinion is not wholly unpredictable. Most of this evidence looks at messages and settings rather than individual people, so it supports the gist of the claim but not the strong word "chaos" without qualification.