{"path":"research/what-a-rule-rests-on/likeness-and-worldview-sources-as-read-2026-10-08.md","content":"# Likeness and worldview: sources as a helper read them (2026-10-08, evening)\n\nA read-only helper (Sonnet; about 16 useful page reads) was asked to test one sentence from\n`docs/worldview-lenses.md` against published work, and to look as hard for evidence against it as\nfor it. Its report is below, unchanged. Afterwards Claude opened one of its sources directly (Braman\nand Nelson 2007, the abstract, from OpenAlex, where it reads cleanly), and it says what the helper\nreports. What Claude made of the report: [../what-a-rule-rests-on.md](../what-a-rule-rests-on.md)\n§ 13.\n\n**The sentence tested:** \"Part of what a worldview does is weigh which likenesses count. Two people\nlook at the same two cases; one sees them as alike in what matters, the other sees a mere detail.\nThat is why a comparison feels apt to one side and forced to the other.\"\n\n**The leads the helper was given** were Claude's, from memory and marked unverified: Tversky 1977;\nGoodman 1972 and 1978; Kuhn on exemplars; Chi, Feltovich and Glaser 1981; Chiu 1972 and Ji, Zhang\nand Nisbett 2004; Braman and Nelson 2007; Khong 1992 and Spellman and Holyoak 1992; Lakoff; Thibodeau\nand Boroditsky 2011 and Steen 2014; Kahan; evidence that likeness follows the verdict; critiques of\nthe moral foundations questionnaire.\n\n---\n\n**Bottom line:** \"Part of what a worldview does is weigh likenesses\" is supported. \"That is why a comparison feels apt to one side\" is the weak part, because the best evidence points to the weighting being at least partly produced by the conclusion.\n\nRead depth key: FT = full text; ABS = abstract only; SEC = secondary summary; SNIP = search-result snippet only (page not opened). Several PDFs came back as binary, so I extracted their text locally with pdftotext (read-only; no files written). That took me past 14 attempts, about 16 of them useful. I stopped there.\n\n| Source · year | Found | Direction | Depth · URL |\n|---|---|---|---|\n| Tversky, Features of Similarity · 1977 | Feature weights depend on context and the object set. \"The relative weights assigned to these features, of course, may differ for different people.\" Grouping comes first and similarity follows it. | Supports; the grouping-first order cuts against \"likeness produces the verdict\" | FT (PDF text layer, reprint) · cseweb.ucsd.edu/~gary/PAPER-SUGGESTIONS/tversky-psych-rev-1977.pdf |\n| \"Revisiting Tversky's diagnosticity principle\" · year not captured | With choice data a 4.75% effect. After controlling for substitution, d = 0.09, p = .26. | Cuts against the diagnosticity hypothesis, not against weights in general | SEC (page via summariser) · pmc.ncbi.nlm.nih.gov/articles/PMC4130183/ |\n| Goodman, Seven Strictures · 1972 | Similarity needs a weighting of importance. \"comparative judgments of similarity often require not merely selection of relevant properties but a weighting of their relative importance\" | Supports | Repost of the essay, extracted by summariser; same wording in a search snippet · pdfcoffee.com/download/goodman-strictures-similarity-pdf-free.html |\n| Goodman, Ways of Worldmaking · 1978 | Confirmed: sub-section \"(b) Weighting\", one of five processes. \"the two worlds contain just the same classes sorted differently into relevant and irrelevant kinds\" | Supports, and shows weighting is only one process | FT (OCR text layer) · archive.org/details/GoodmanWaysOfWorldmaking |\n| Medin, Goldstone & Gentner · 1993 | Similarity is flexible, but \"respects\" are fixed by the comparison process, not by free-floating weights. Knowledge decides which features are constructed, not only how they are weighted (citing Wisniewski & Medin 1991). Quotes Murphy & Medin: similarity is \"at least as much a consequence as a cause of conceptual coherence\". | Both | FT, read selectively; OCR garbled · groups.psych.northwestern.edu/gentner/papers/MedinGoldstoneGentner93.pdf |\n| Nosofsky · 1986 | Model with attention weights that set how much each attribute contributes to similarity; lab stimuli | Supports narrowly (lab dimensions, not worldviews) | SNIP · psych.indiana.edu/documents/nosofsky-1986.pdf |\n| Chi, Feltovich & Glaser · 1981 | Novices sort physics problems by surface features, experts by principle | Supports. The cause is knowledge, not motive. | SEC (as reported in the MGG93 text) |\n| Kuhn: SEP entry; \"Second Thoughts\"; Postscript | Scientists judge new problems by similarity to exemplars, and perception of similarity is learned and not rule-governed. Incommensurability involves changes in meanings and taxonomies, not just re-weighting. | Both | SEC (SEP, via summariser; no quote offered) · plato.stanford.edu/entries/thomas-kuhn/ |\n| Le, Gao, Frank & Carstensen · 2023 (year from filename) | Replicates the US–China difference in thematic versus taxonomic similarity. Vietnam falls in between. \"Perhaps notions of similarity are similar across contexts, but the statistics of the linguistic environment vary.\" Their intro summarises Chiu 1972 and Ji et al. 2004. | Both | FT of abstract and intro · abcarstensen.com/files/LeGaoFrankCarstensen23_Reexamining-crosscultural-similarity-judgments.pdf |\n| Ji, Zhang & Nisbett · 2004 | Chinese bilinguals grouped more relationally than European Americans, in either test language | Supports | SNIP only |\n| Braman & Nelson, AJPS · 2007 | Two experiments (undergraduates, then undergraduates plus law students). Participants rated cases whose outcomes fit their policy views as more analogous. The effect was largest when cases were moderately similar, and legal training did not reduce it. \"Attitudes influence judgments by affecting the perceived similarity between a target case and cases cited as precedent.\" | Supports, and the direction runs from attitude to likeness | ABS (OpenAlex-reconstructed abstract, garbled in places) · api.openalex.org/works/doi:10.1111/j.1540-5907.2007.00290.x |\n| Kahan et al., \"They Saw a Protest\" · 2012 | Subjects holding the same belief about the protest but opposed cultural outlooks disagreed sharply on key facts. Named mechanisms: crediting information selectively, imputing expertise to the like-minded, and being \"more likely to note, assign significance to, and recall facts supportive of their cultural outlooks\". | Both: identity-driven perception, but about facts and trust, not likeness | FT (abstract and theory section; sample size not read) · stanfordlawreview.org/wp-content/uploads/sites/3/2012/05/Kahan-64-Stan-L-Rev-851.pdf |\n| Simon et al. 2001, reporting Holyoak & Simon 1999 | Ratings of legal arguments shift toward coherence with the emerging verdict: \"the emerging decisions worked backwards to alter the strength of the inferences\". The 2001 study found shifts even before any decision. These are argument ratings, not similarity ratings. | Cuts against verdict-independent weighting | FT of abstract and intro (the 1999 paper not opened) · reasoninglab.psych.ucla.edu/wp-content/uploads/sites/273/2021/04/Simon_etal1.2001.pdf |\n| Steen, Reijnierse & Burgers · 2014 | Four experiments (N = 253, 259, 252, 1,026) found no beast-versus-virus framing effect. \"Reading about crime increases people's preference for enforcement irrespective of metaphorical frame\" | Cuts against metaphor as a likeness-chooser | FT · journals.plos.org/plosone/article?id=10.1371/journal.pone.0113536 |\n| Harper & Rhodes · 2021 | Only three clusters emerged from the Moral Foundations Questionnaire, not five or six. \"only three meaningful clusters emerged in an exploratory principal factors analysis\" | Cuts against the questionnaire's measurement | ABS (via Semantic Scholar API) · api.semanticscholar.org/graph/v1/paper/DOI:10.1111/bjso.12452 |\n| Spellman & Holyoak · 1992; Khong · 1992 | Gulf War–WWII analogy mapped \"bistably\" into two coherent sets. Khong: analogies used to make decisions, not just justify them. | Unclear; Khong would favour likeness → verdict | SNIP only |\n\n**Verdict**\n\n- **Weights are supported.** Similarity depends on weighted respects (Tversky, Goodman, MGG93). Groups differ in them: experts versus novices, policy camps (Braman & Nelson), and partly cultures.\n- **Weights are only part of the story.** Goodman lists weighting as one of five worldmaking processes. MGG93 says knowledge changes which features exist, and the SEP summary stresses meaning and taxonomy change. Kahan's mechanisms are about facts and trust. \"A weighting\" alone is too thin as a full description of a worldview.\n- **The weights are partly produced by the conclusion.** Braman & Nelson's participants found cases more alike when the outcome suited their views. Argument strength shifts toward the emerging verdict (Holyoak & Simon). Tversky himself has similarity following grouping. Chi's expert effect is the exception, and it comes from knowledge, not wanting a verdict.\n- **Suggested wording.** Say a worldview tilts which likenesses count, and that the tilt is partly set by the conclusion already wanted. \"That is why\" overstates.\n\n**Leads that did not check out or stayed unread**\n\n- Feinberg et al. 2020 on Lakoff: search result only. It tests whether family-model values predict political attitudes, not likeness weighting, so it is off-target. The snippet's author list is Feinberg, Wehling, Chung, Saslow and Paulin.\n- Thibodeau & Boroditsky 2011: not opened. Steen's title says it follows up their 2013 paper, so check which paper you mean.\n- Graham, Haidt & Nosek 2009: not opened.\n- Kuhn Postscript date: the SEP summary said 1970, and I could not check your 1969.\n- Kuhn 1977 on shared theory-choice values applied with different weights: from memory, unverified. It may be the closest Kuhn text to a literal weighting account.\n- Ji et al. replication status: the 2025 review in the Journal of Pacific Rim Psychology style (Springer s41809-025-00166-w) was blocked.\n- Collabra 2026 MFQ-1 paper: 403.\n- Murphy & Medin 1985, Wisniewski & Medin 1991, Kunda 1990: seen only as citations in the texts above.\n"}