{"id":"e7d4d86e-d8bf-4922-bdbc-1729a8a46666","arxiv_id":"2602.22395","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Modern misinformation research inherits its core model from 1980s Satanic-panic-era memory research, not just from social media.","lead":"Using 13,339 academic papers from 2011–2023, this study traces the word \"misinformation\" and argues that today's fake-news research inherits its core model from 1980s Satanic-panic memory science, not just from social media. The paper matters because it rewrites the origin story of a major research field and highlights assumptions about individual belief versus media structure.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Pre-2016 term shifts attributed to the Loftus community are never shown to originate there; a distributional test is needed before the lineage claim is established.","rationale":"The reader’s weakest_assumption identifies the same gap: the paper treats pre-2016 frequency shifts as evidence of the Loftus community’s influence without a statistical test linking those words to that community, and it designates #1432 as the only well-established pre-2016 strand using a top-community selection rule. My concern is the same, sharpened to the single step that most directly supports the central claim: the attribution of the pre-2016 linguistic signal to the Loftus community. This is load-bearing because if those words are not concentrated in #1432, then the paper’s evidence for a distinctive pre-2016 lineage collapses, and the claim that the post-2016 paradigm ‘owes more’ to it than acknowledged has no empirical basis beyond a few citations and conceptual parallels. The authors’ Section D limitation statement explicitly acknowledges the single-keyword dependence and the possible existence of other strands, which supports a conditional rather than a rejection verdict. The suggested test would settle whether the term-shift evidence actually lands. The reader’s conditional verdict is appropriate; I would keep it conditional pending that test, so no adjustment to the verdict is needed.","tokens_in":16818,"tokens_out":5898,"duration_ms":62224,"concrete_test":"Re-run the Section II.A term-frequency analysis separately for papers in community #1432 and for all other pre-2016 papers. For each of the nine largest pre-2016 shift terms (memory, women, children, recall, etc.), compute a log-odds ratio with a Dirichlet prior and a permutation test over community labels, with multiple-comparison correction. Also compute the fraction of all pre-2016 occurrences of each term that come from #1432. If these terms are not significantly enriched in #1432, or if #1432 contributes less than a clear majority of their occurrences, then the assertion that #1432 “explains” the pre-2016 frequency shifts is unsupported, and the central lineage claim must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that post-2016 misinformation research owes more to the Loftus/‘misinformation effect’ lineage than acknowledged—depends on the assertion in Section III that the Loftus community (#1432) is “the only well-established pre-2016-paradigm strain of misinformation research, one that explains the pre-2016 frequency shifts in terms.” That assertion is not backed by a distributional test. The pre-2016 shift terms in Figure 3 (memory, women, children, recall, etc.) are computed over the full Scopus corpus, not over community #1432. Nothing in the paper shows that these terms are concentrated in #1432 relative to all other pre-2016 papers, or that #1432 accounts for a substantial share of their pre-2016 occurrences. The top TF-IDF terms for #1432 in Table I (ethical, debriefing, deception, researchers, dialogue) are not the same as the shift terms, and the single illustrative abstract from Otgaar et al. is not evidence of corpus-level attribution. Because words like “women,” “children,” “memory,” and “recall” are common in health, psychology, and legal contexts that use “misinformation” incidentally, their pre-2016 frequency could reflect non-Loftus fields. The paper’s Section D limitation candidly concedes that a single keyword and possibly unidentified pre-2016 strands undermine the strongest reading of the claim. The limitation is honest, but the headline conclusion is stronger than the reported evidence permits.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the evolution of the term \"misinformation\" in academic literature by analyzing a Scopus export of papers with \"misinformation\" in any metadata field from 2011 to 2023. It identifies pre-2016 and post-2016 term-frequency shifts, constructs author-sharing networks with Louvain community detection, and finds a pre-existing community (#1432, the \"Loftus community\") centered on the misinformation effect and false-memory research. The authors argue that this lineage, which they trace to the Satanic panic of the 1980s, is the only well-established pre-2016 scientific strand of misinformation research and that the post-2016 social-media-focused paradigm inherits its core model—individuals receive false content, memory/belief is corrupted, and interventions correct it—from that source. They discuss theoretical commitments, simplifying assumptions, and parallels between the Satanic panic and contemporary misinformation crises, concluding with limitations (single keyword, proprietary data, single data source).","tokens_in":17194,"tokens_out":6700,"duration_ms":56907,"significance":"If the lineage claim held, the paper would be a valuable corrective to standard histories that date misinformation research to social media and the 2016 election. It would show that a central cognitive model predates the current paradigm and that its origins are entangled with the Satanic panic, with implications for how the field frames its problems. The paper is clearly written, historically engaged, candid about its limitations (Section D), and does not argue circularly: its identification of the Loftus community is grounded in author-sharing and external citation examples. However, the empirical support is exploratory; the central attribution is not yet established by the data presented. The paper's contribution, if the recommended distributional tests are added, would be a useful historical and conceptual analysis rather than a definitive bibliometric demonstration.","major_comments":[{"comment":"The claim that #1432 is 'the only well-established pre-2016-paradigm strain ... one that explains the pre-2016 frequency shifts in terms' is the load-bearing step, but is unsupported. The shift terms in Fig. 3 (memory, women, children, recall, patients, event) are computed over the full corpus, not conditionally on #1432. No evidence shows these terms are concentrated in #1432 or that #1432 accounts for a substantial share of their occurrences. Table I lists different top TF-IDF terms for #1432, and the Otgaar et al. abstract is anecdotal. These words are common in health/psychology/legal papers that use 'misinformation' incidentally, so the shifts may reflect non-Loftus strands. Report a distributional test (e.g., log-odds of each shift term in #1432 vs. the rest, with uncertainty, and the proportion of pre-2016 occurrences in #1432) before sustaining the attribution.","section":"Section III, Fig. 3, Table I"},{"comment":"The conclusion that no other pre-2016 strand was 'well-established' rests on selecting the two largest communities per year and on Louvain resolution. Of 8033 communities, 6797 are singletons; small but coherent pre-2016 strands could be split or absorbed by the algorithm. No sensitivity analysis is given (different resolution, author-level network, or a size threshold for pre-2016 communities). Section D admits other strands, but the 'only' in Section III is stronger than the evidence permits. A robustness check or a more limited claim is needed.","section":"Section II B, Fig. 4(a)"},{"comment":"The term-shift analysis selects the 9 most extreme positive and negative frequency differences and a bigram threshold of >100, with no sensitivity analysis. The narrative that pre-2016 terms are 'unexplained' depends on which tokens survive these thresholds; please show stability across k and threshold, or at least report how Figure 3's word lists change.","section":"Section II A, Fig. 3"}],"minor_comments":[{"comment":"Define 'metadata' precisely and clarify that the bag-of-words analysis uses titles and abstracts only, while corpus selection used all metadata fields.","section":"Section II A"},{"comment":"Spell out how shared authorship is counted in the edge weight, especially when two papers share multiple authors.","section":"Section II B"},{"comment":"'Two biggest communities of each year' is ambiguous; clarify whether selection is by annual community size and why 'all the biggest communities in 2023' equals the top 10.","section":"Figure 4 caption"},{"comment":"Typos: 'complimentary' should be 'complementary' in both occurrences; 'Alshaabiet al.' is malformed in the Figure 1 caption.","section":"Sections III and IV A"},{"comment":"The word 'nihilogony' appears without definition; consider a gloss or replacement.","section":"Section IV A"},{"comment":"Clarify that the Twitter data come from Alshaabi et al. and are used only as an illustration, since the main analysis is restricted to Scopus.","section":"Figure 1"}],"recommendation":"major_revision","confidential_remarks":"The paper is likely to be of interest to an interdisciplinary audience, and the historical narrative is valuable. The main risk is that the abstract and Section III make a stronger attribution than the reported evidence supports. I would not reject; the authors should either add a distributional attribution test tying the pre-2016 shift terms to community #1432, or substantially soften the 'only well-established' wording relative to their own Section D limitations."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper argues that the post-2016 misinformation paradigm's core model—individuals receive false content, it corrupts belief or memory, and interventions correct it—comes from Loftus-era memory research rooted in the Satanic panic, and that standard histories starting in 2016 are incomplete. That claim is fresh and worth taking seriously. It's not in the cited histories, and the paper does a real service by pointing to the pre-2016 \"misinformation effect\" literature and showing how it gets folded into the social-media era.\n\nWhat the paper does well: the historical synthesis is thoughtful, the limitations section is unusually honest (single keyword, proprietary Scopus data, possible missing strands), and the examples (Otgaar et al., Loftus's own 2016 election paper) make the lineage concrete. The conceptual discussion of why this matters—the field's simplifying assumptions about misinformation as link-sharing, the neglect of media structure—is the strongest part and stands on its own.\n\nWhere it's soft: the empirical support for the central attribution is exploratory. The pre-2016 frequency shifts for \"memory,\" \"women,\" \"children,\" and \"recall\" are computed over the whole corpus, not over the Loftus community. The paper asserts that community #1432 is \"the only well-established pre-2016-paradigm strain\" and that it \"explains\" those shifts, but there's no distributional test showing those terms are concentrated in #1432, or that #1432 accounts for a substantial share of pre-2016 occurrences. The top TF-IDF terms for #1432 (ethical, debriefing, deception) are not the same as the shift terms. So the stress-test concern lands: the headline conclusion is stronger than the evidence. The authors acknowledge this in Section D, but the abstract's \"owes more than is generally acknowledged\" is still a reach given the current analysis.\n\nThat said, the paper is not circular, the self-citations are incidental, and the argument is coherent on its own terms. It deserves a serious referee: the historical thesis is important, the writing is clear, and the empirical gaps are fixable (release the corpus and code, add a formal attribution test, check community resolution robustness). I'd send it to review, expecting revision rather than rejection.\n\nWho it's for: misinformation researchers, science historians, and anyone who wants a bracing corrective to the field's short memory. I'd bring it to reading group and would cite it as a provocative counter-narrative, with a caveat about the evidence.","headline":"Plausible and thought-provoking genealogical claim that post-2016 misinformation research inherits its core model from 1980s memory science, but the corpus evidence is exploratory and doesn't yet nail the attribution.","tokens_in":17661,"tokens_out":1609,"would_cite":true,"duration_ms":15374,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Post-2016 misinformation research inherits its core model from 1980s false-memory science that emerged during the Satanic panic, not just from social media.","keywords":["misinformation","misinformation effect","false memory","Satanic panic","history of science","community detection","term frequency","social media misinformation"],"falsifier":"If a separate substantial pre-2016 community of papers (for example, in political communication or public health) used \"misinformation\" and shaped post-2016 work, or if the pre-2016 frequency of terms like \"memory\" and \"recall\" is spread across many communities rather than concentrated in the false-memory lineage, the central claim would be weakened.","tokens_in":16748,"feed_emoji":"🧠","tokens_out":4474,"duration_ms":47430,"temperature":0.7,"pith_summary":"The paper argues that the explosion of misinformation research after 2016 is not simply a new paradigm born of social media and the 2016 election. Instead, it claims the field's central model—individuals receive false content, it corrupts belief or memory, and interventions correct it—comes from an older strand of psychological research on the misinformation effect. That strand grew out of scientific responses to the Satanic panic of the 1980s, when recovered-memory therapy and coercive interviewing were producing false memories in legal cases. By tracing the term \"misinformation\" through academic literature, the authors connect today's social-media-focused research to that earlier memory-science lineage and argue that standard histories are incomplete.","feed_headline":"Misinformation science grew out of 1980s false-memory research","feed_subtitle":"A bibliometric study finds the post-2016 social-media paradigm inherits its core model from the Satanic-panic era.","key_machinery":"The central mechanism is a bibliometric genealogy built from an author-sharing network of 13,339 papers with the word \"misinformation\" in their metadata from 2011 to 2023. Edges between papers are weighted by the inverse of the product of their author counts so that papers with many authors do not dominate. A community-detection algorithm partitions the network into research clusters; the largest stable pre-2016 cluster is the false-memory and misinformation-effect community, while other clusters are tiny until after 2016. Comparing term frequencies before and after 2016 surfaces words like \"memory,\" \"recall,\" \"women,\" and \"children\" that are unexplained by social-media-centric histories, an","core_discovery":"The paper's central claim is that the post-2016 misinformation paradigm owes more to the 1980s misinformation-effect research lineage than is generally acknowledged. Tracking papers that contain \"misinformation\" in their metadata, the authors identify a stable pre-2016 community of memory-science papers built around experiments showing that exposure to misleading post-event information can distort memory. This community predates social-media misinformation research, was the largest well-established pre-2016 strand, and is now being absorbed into the modern paradigm through concepts like prebunking and inoculation. The paper reads this lineage as rooted in scientific responses to the Satanic","pith_inferences":["If the lineage claim is right, post-2016 correction interventions may be implicitly testing a memory-suppression model; a direct comparison with platform-design interventions would reveal whether that focus is a limiting choice.","The keyword-based method could be extended to \"disinformation\" and \"fake news\" corpora to see whether those terms share the same false-memory ancestry or have different intellectual roots.","The paper's boundary-work framing suggests a measurable prediction: misinformation papers will disproportionately treat platform companies and audiences as objects to be corrected rather than as interlocutors whose design choices shape the problem.","One could examine citation patterns over time: if the lineage thesis is correct, post-2016 papers should increasingly cite memory-science work when introducing \"misinformation,\" even outside psychology venues."],"forward_implications":["Accepted histories of misinformation that begin with social media and 2016 are incomplete; the field's intellectual roots extend to 1980s memory science.","The modern paradigm's focus on individual content, sources, and consumers—rather than media structure or political economy—is a legacy of the lab-based misinformation-effect model.","Intervention concepts like prebunking and inoculation are direct descendants of older techniques for protecting memory from post-event misinformation.","The parallel between the Satanic panic and the current QAnon-era panic suggests a recurring pattern in how science responds to moral panics, including scientists acting as boundary-keepers of epistemic authority.","Defining misinformation as posts linking to unreliable domains inherits the older model's simplifying assumptions, which fit awkwardly with the scale and design of social-media platforms."],"fun_headline_variants":["Misinformation science grew from 1980s panic research","False-memory studies of the 80s birthed fake news science","Today's misinformation paradigm traces to Satanic panic era","Modern misinformation research owes a debt to 1980s memory science"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The entire lineage claim rests on treating a keyword search for \"misinformation\" and the largest author-sharing communities as a faithful map of the field's intellectual history; if the keyword or the community partition leaves out other substantial pre-2016 strands, the claimed ancestry does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Misinformation science grew from 1980s panic research","False-memory studies of the 80s birthed fake news science","Today's misinformation paradigm traces to Satanic panic era","Modern misinformation research owes a debt to 1980s memory science"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000194,"raw_usage":{"total_tokens":1147,"prompt_tokens":658,"completion_tokens":489,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":402,"completion_tokens_details":{"reasoning_tokens":417}},"tokens_in":402,"tokens_out":489,"duration_ms":5490,"temperature":1.0,"reasoning_tokens":417,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T20:42:34.176312+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If a separate substantial pre-2016 community of papers (for example, in political communication or public health) used \"misinformation\" and shaped post-2016 work, or if the pre-2016 frequency of terms like \"memory\" and \"recall\" is spread across many communities rather than concentrated in the false-memory lineage, the central claim would be weakened.","supporting_citations":[],"review_version":1}