Two neighbours argue about a new parking rule. One says the rule is greedy. The other says the rule is fair. They talk for an hour and neither moves. A week later, one of them finds out what the other actually meant. It was not the thing she spent the hour beating.
She won. She also never met the argument.
Sources
Brand, C. O., Brady, D., & Stafford, T. (2025). The Ideological Turing Test: A Behavioral Measure of Open-Mindedness and Perspective-Taking. Cognitive Science 49(10), e70126. Methods re-read this pass and the module CORRECTED against them: TWO separate samples — 600 participants GENERATED arguments (100 per side per topic; three for their own side, three for the other), and a DIFFERENT 1,200 participants RATED them (200 per side per topic), for 1,800 in all. Prolific, UK-based, 'three topics with potential for polarization in the UK of the early 2020s'. Raters were blind: 'participants were not told whether the arguments they were rating were written by people who were for or against a particular topic'; each rated 60 arguments, each argument rated by 10 people per side. Pass rates in the module are the RELATIVE criterion (opponent ratings equal to or above the mean of proponents' own arguments): 71% veganism, 64% Brexit, 54% vaccines. NOTE the paper also reports an ABSOLUTE criterion on which non-vegans (71%) beat vegans (32%) — the module's 'neither side was better' claim is therefore qualified to 'on the main measure'. Novelty claim verified: 'To our knowledge, no one has operationalized the ITT to allow formal assessment...'. Covid arguments were re-used from Brand & Stafford (2022), so the study was run before its 2025 publication. https://pmc.ncbi.nlm.nih.gov/articles/PMC12519043/
Caplan, B. (2011). The Ideological Turing Test. EconLog, 20 June 2011. Original post fetched; the two quoted sentences are VERBATIM and complete (the module truncates before the symmetrical Krugman half, whose context the following body block supplies). Date confirmed. He was answering Krugman's claim that one side can simulate the other and the other cannot — the module deliberately does not name the sides, which keeps it non-partisan. https://www.econlib.org/archives/2011/06/the_ideological.html
Messinger, C. (2012). Knocking Down a Steel Man: How to Argue Better. The Merely Real, 7 December 2012. Original post fetched; the quoted clause is VERBATIM ('steelmanning, the art of addressing the best form of the other person's argument, even if it's not the one they presented'). Title and date confirmed. https://themerelyreal.wordpress.com/2012/12/07/steelmanning/
Dennett, D. C. (2013). Intuition Pumps and Other Tools for Thinking. Rule 1 confirmed VERBATIM against a reproduction of the book passage, including the internal quotation. All four rules confirmed in order. IMPORTANT provenance correction applied to the module: the numbered four-rule list is DENNETT'S synthesis of Rapoport, not a list Rapoport published as four numbered rules — the sequence_builder prompt previously said 'the order he gave them' (Rapoport) and now says Dennett's. Secondary accounts trace the rules to Rapoport's 1960 book plus his 1961 paper 'Three Modes of Conflict', with the first rule credited by Rapoport himself to Carl Rogers; that provenance detail is NOT asserted in the module because it rests only on secondary sources. https://www.themarginalian.org/2014/03/28/daniel-dennett-rapoport-rules-criticism/
Rapoport, A. (1960). Fights, Games, and Debates. University of Michigan Press. ⚠️ THE BOOK ITSELF WAS NOT READ — publisher listing plus secondary summaries only. What those support: fights aim to eliminate the opponent, debates aim to convince a rational opponent rather than destroy them (the module's 'In a fight you remove your opponent. In a debate you have to reach them.'), and Rapoport's own line that 'conveying assurance of understanding to the opponent is the first step in the removal of threat', which backs 'a debate opens by stating the other side's case'. Publication year and publisher confirmed against the American Political Science Review and Science review listings. https://press.umich.edu/Books/F/Fights-Games-and-Debates
Word Spy, 'steel man'. Earliest citation confirmed VERBATIM: 8 August 2003, Paul C. Anagnostopoulos, James Randi Educational Foundation forum — 'You know, this straw man is so big and so strong that he'll soon morph into a steel man.' The module paraphrases rather than quotes (it drops 'and so strong'), which is why it is not in a quote block. https://wordspy.com/words/steel-man/
Eyal, T., Steffel, M., & Epley, N. (2018). Perspective mistaking: Accurately understanding the mind of another requires getting perspective, not taking perspective. JPSP 114(4), 547-571. ABSTRACT READ VERBATIM from the authors' own PDF this pass, and two module claims corrected against it. (1) Domains are 'predicting another person's emotions from facial expressions and body postures, predicting fake versus genuine smiles, predicting when a person is lying or telling the truth, and predicting a SPOUSE'S ACTIVITY PREFERENCES AND CONSUMER ATTITUDES' — the module's 'spouse's taste in films' was invented and has been replaced. (2) The confidence finding is hedged in the original: 'If anything, perspective taking decreased accuracy overall while OCCASIONALLY increasing confidence in judgment' — the module's flat 'It did raise confidence' has been softened. 25 experiments confirmed. Final experiment: 'getting another person's perspective directly, through conversation, increased accuracy'. https://www.nicholasepley.com/publications
Greater Good Magazine (UC Berkeley). How Putting Yourself in Someone Else's Shoes May Backfire. Used only to confirm that the final Eyal et al. experiment had participants ask direct questions. https://greatergood.berkeley.edu/article/item/how_putting_yourself_in_someone_elses_shoes_may_backfire
Fernbach, P. M., Rogers, T., Fox, C. R., & Sloman, S. A. (2013). Political Extremism Is Supported by an Illusion of Understanding. Psychological Science 24(6), 939-946. Confirmed: mechanistic explanation both undermined the illusion of explanatory depth and moderated attitudes (Exps 1-2), and critically 'these effects occurred when people were asked to generate a mechanistic explanation, but they did not occur when people were instead asked to enumerate reasons' (Exp 2) — which is exactly what the module's reveal says. https://journals.sagepub.com/doi/abs/10.1177/0956797612464058
Crawford, J. T., & Ruscio, J. (2021). Asking People to Explain Complex Policies Does Not Increase Political Moderation: Three Preregistered Failures to Closely Replicate Fernbach, Rogers, Fox, and Sloman's (2013) Findings. Psychological Science 32(4), 611-621. Abstract read. Confirmed: three preregistered replications; 'policy-position extremity and in-group favoritism were unaffected'; BUT 'participants reported less policy understanding after providing mechanistic policy explanations' — so the drop in self-rated understanding DID replicate while the moderation did not. The module states both halves correctly. https://pubmed.ncbi.nlm.nih.gov/33667138/
Sloman, S. A., & Vives, M.-L. (2022). Is political extremism supported by an illusion of understanding? Cognition 225, 105146. Abstract read. Confirmed: reduced extremity on consequentialist issues, no reduction on issues held as protected values. Attribution corrected in the module from 'the original lab' to 'one of the original authors' — Sloman is one of the four 2013 authors, who were spread across several institutions. https://pubmed.ncbi.nlm.nih.gov/35533417/
memory/content/curriculum/critical-thinking.md — module 5.5 scope ('Charity and steelmanning, done properly — and its limits, including the failure mode where a steelman is a strawman with better manners because you built the version YOU would find convincing'), and the Pathway 5 instruction to keep the material scrupulously non-partisan. The module hits the named failure mode squarely in 'Strongest and honest' and 'Who the tool is pointed at'.
memory/content/wave2/productive-conflict-v2-REVIEW.md §9 — confirms straw man, steelmanning and principle of charity are already glossed there via tap_to_define, and that the 'restating is not agreeing' reveal is already spent. This module correctly refers to straw man in one sentence and never re-teaches it, and never re-teaches the principle of charity.
packages/db/src/content.ts and apps/web/src/components/blocks/*.tsx — renderer conformance checked in code. All 11 block types used are handled on both surfaces; every field validates. SortIntoBuckets defaults doShuffle = true (items shuffle on mount), which is what made section 11's 'Look at the third item again' a defect; MatchPairs shuffles only the right column; SequenceBuilder shuffles and therefore expects steps authored in correct order, which they are.