{"id":"798f54c0-c821-4ab4-9483-e2313c90eca6","arxiv_id":"2412.04999","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Health professionals on TikTok debunk nutrition and mental health misinformation through a shared five-step process driven by perceived harm, scientific evidence, and symmetric duet-style responses.","lead":"Fourteen health professionals who debunk TikTok health misinformation describe a common process: seeing a viral false claim, picking which ones to counter, and stitching a science-based response video to the original. The paper maps this do-it-yourself debunking model and contrasts it with platform warning labels and community notes.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Recruitment via #duet/#stitch hashtags selects on the very response format the process model claims to discover, making the 'symmetric response' and 'stitching/duetting' stage potential sampling artifacts.","rationale":"The reader's weakest assumption identifies the #duet/#stitch hashtag sampling premise as load-bearing and worries that debunkers who do not duet or stitch were systematically missed. My concern sharpens this: it is not merely an incompleteness risk but a selection-on-outcome circularity. Because the population frame was constructed using #duet/#stitch, the paper's central claim that the counterinfluence process 'follows a common process of ... stitching or duetting' and that it 'offers a symmetric response' is predetermined by the inclusion criteria. No alternative debunking format could have appeared in the sample, so the unique-aspect claims lack comparative support. This reinforces the reader's CONDITIONAL verdict: the findings should be explicitly framed as describing duet/stitch-based health professional debunkers, not all health professional debunkers on TikTok. I do not think the concern requires rejection, since the study is exploratory and the process stages other than stitching/duetting (initiation, selection, creation, post-debunking management) were not directly selected on, but the abstract and Section 6 language overgeneralizes and should be qualified.","tokens_in":32032,"tokens_out":4103,"duration_ms":42832,"concrete_test":"Report the recruitment funnel: of the 135 contacted accounts, how many were found through queries containing #duet or #stitch versus queries without those hashtags? If all or nearly all 135 were identified via #duet/#stitch, the process model's symmetric-response stage is a selection criterion. Stronger test: re-run the Section 4.1 recruitment on debunking videos sampled from #adhd/#anorexia without filtering on #duet/#stitch, manually vet for debunking content, and measure the proportion of creators whose primary response format is duet/stitch. If that proportion is materially below 100%, the claimed 'common process' is an artifact of the sampling frame.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim asserts a common DIY process culminating in 'stitching or duetting a debunking video with a misinformation video' and a unique 'symmetric content-to-content response' (Abstract, Sections 5.4, 6). But Section 4.1 states that to discover debunking videos, the researchers 'applied the hashtags #duet and #stitch in these content areas.' The recruitment frame therefore requires that every sampled debunker has produced at least one duet or stitch video. The subsequent finding that participants use duets/stitches as their dominant tactic (Table 6, 9 of 14 codes) is guaranteed by the inclusion criterion, not empirically discovered. This is a selection-on-outcome problem: the process model's defining stage and the 'symmetric response' unique aspect are built into the sampling design. The comparison to asymmetric moderation (labels, community notes) is not a comparison to other debunking formats (original videos, comments, text posts) because those formats were excluded from the population frame. The paper's limitation section (6.5) does not acknowledge this circularity, so the abstract's universal phrasing overstates what the data can support: at most, the study describes health professionals who debunk using duet/stitch affordances, not health professional TikTok debunkers generally.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents an exploratory qualitative study of 14 health professionals who debunk nutrition and mental health misinformation on TikTok by creating counter-videos. Using the TikTok API, the authors recruited participants through hashtags including #duet and #stitch, administered a survey, and applied thematic analysis. They propose a five-stage 'Debunk-It-Yourself' process: initiation, selection, creation, responding via duet/stitch, and post-debunking management. They further claim three unique aspects: targeting trending harmful misinformation, offering a symmetric content-to-content response, and grounding responses strictly in scientific evidence and claimed clinical experience. The paper contrasts this effort with platform warning labels and community notes and offers recommendations.","tokens_in":32380,"tokens_out":6900,"duration_ms":250362,"significance":"The topic is timely and the study is one of the first to examine the 'supply side' of grassroots, professional-led debunking on a short-video platform. The paper's strengths include a detailed codebook (Appendix B), the full survey instrument (Appendix A), explicit ethical safeguards, and a transparent recruitment and vetting procedure. The inter-rater reliability (Cohen's kappa = 0.8334) is acceptable for qualitative coding. If the central process model were established on a representative sample, this would be a meaningful contribution to the misinformation literature. However, the recruitment strategy selects on the very response format (duet/stitch) that the model claims to discover, which materially threatens the universal phrasing of the central claim. The paper is better viewed as an account of health professionals who debunk through duet/stitch affordances, and the claims need to be qualified accordingly.","major_comments":[{"comment":"The recruitment procedure in Section 4.1 states: 'To discover videos debunking misinformation, we applied the hashtags #duet and #stitch in these content areas.' This means the population frame is restricted to debunkers who have used duet or stitch for at least one debunking video. Consequently, the abstract's claim that the counterinfluence 'follows a common process of initiation, selection, creation, and \"stitching\" or duetting a debunking video with a misinformation video' is partly guaranteed by the inclusion criteria, not discovered from the data. Similarly, the 'symmetric content-to-content response' presented as a unique aspect (Section 6) is a consequence of sampling from duet/stitch videos. This is a selection-on-outcome problem. The authors should either reframe all conclusions as applying specifically to health professionals who debunk via duet/stitch, or broaden recruitment to include debunkers using other formats (e.g., original videos, text overlays, comments) and then test whether the five-stage process still holds.","section":"Section 4.1; Table 6; Abstract"},{"comment":"Table 6 reports only 9 of 14 participants using duets/stitches as an 'own influencing tactic,' while the recruitment in Section 4.1 implies every participant must have produced at least one duet/stitch debunking video to be in the sample. The authors need to explain this discrepancy. Possibilities include that Table 6 codes a self-reported primary tactic rather than any use of the affordance, or that the hashtag-based discovery did not strictly require the creator's own video to carry #duet/#stitch. Without clarification, the empirical support for the 'common process' claim is ambiguous: the claim may hold for a minority of the sample even though the sampling frame would predict a majority.","section":"Table 6 vs. Section 4.1"},{"comment":"The limitation section acknowledges sample size, English-language sampling, and lack of impact assessment, but it does not acknowledge the selection-on-outcome issue in the recruitment hashtags. This omission is load-bearing because the process model's final stage and the 'symmetric response' unique aspect are directly affected. The authors should add an explicit discussion of how the #duet/#stitch-based sampling shapes the findings and qualify the abstract's universal wording (e.g., 'among health professionals who debunk through duet/stitch').","section":"Section 6.5"},{"comment":"The quantitative analysis reports that #stitch is the third most-used hashtag among the examined debunkers and interprets this as evidence that stitching is the main affordance these users leverage. Because the population was discovered by applying #duet and #stitch as search hashtags, this observation is circular. The hashtag distribution cannot be used to infer the relative importance of duet/stitch without a sample not selected on those hashtags. The authors should either re-analyze the hashtag distribution from a non-circular sample or remove this claim.","section":"Section 5.10; Figure 5"}],"minor_comments":[{"comment":"Typo: 'for for their participation' should be 'for their participation'.","section":"Section 4.1"},{"comment":"Typo: 'familiary' should be 'familiar'.","section":"Section 5.5"},{"comment":"Typo: 'therms' should be 'terms' in the quote from P2.","section":"Section 5.2"},{"comment":"The text refers to the 'Disinforamation Handbook' in the description of P6's response; this should be 'Debunking Handbook' to match reference [44].","section":"Section 5.9"},{"comment":"Typo: 'beleive' should be 'believe'.","section":"Section 6.5"}],"recommendation":"major_revision","confidential_remarks":"The paper has a clear and significant potential contribution, but the sampling-on-outcome issue is not merely a limitation; it directly shapes the central process model and the 'symmetric response' claim. The authors should be given the opportunity to revise by explicitly restricting the claims to duet/stitch-based debunkers or by providing additional evidence from a broader sample. The paper's current abstract overstates the universality of the process. The detailed codebook and ethical treatment of participants are commendable and should be preserved in revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the five-stage process model—initiation, selection, creation, response, post-debunking management—plus the specific selection criteria (virality, blatant falsehood, susceptibility) and the deliberate avoidance of third-party fact-checkers. The codebook is detailed, the survey instrument is in the appendix, and the authors vetted credentials rather than trusting avatars. The IRR is acceptable. That is real descriptive value for a subfield that mostly studies labels and comments.\n\nThe soft spot is structural. Section 4.1 says they found debunking videos by applying the hashtags #duet and #stitch. Every participant therefore produced at least one duet or stitch video by construction. Table 6's finding that duets/stitches are the dominant tactic, and the 'symmetric content-to-content response' in the abstract, are selection artifacts. The paper does not acknowledge this in Section 6.5, and the abstract's universal phrasing (\"follows a common process\") goes beyond what a hashtag-restricted sample can support. At most, this describes health professionals who debunk through duet/stitch, not health professional TikTok debunkers generally. That is a load-bearing caveat, not a nit.\n\nThere is also a smaller but real data-handling issue: Section 5.5 quotes participant P16 in a study with n=14. That needs a fix and a check of the other participant labels. Minor spelling issues aside, the rest of the analysis is careful.\n\nMy take: the process model is plausible for the sampled subpopulation, but the paper overgeneralizes. An editor should send it to peer review with a clear request to reframe the contribution, acknowledge the selection-on-outcome, and soften the claims. The phenomenon is under-documented and this is a reasonable first step, but it is not the definitive account the abstract suggests.","headline":"A useful first map of duet/stitch debunking by health professionals, but the sampling design guarantees the headline behavior, so the process model outruns the data.","tokens_in":32775,"tokens_out":3705,"would_cite":true,"duration_ms":34608,"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":"Health professionals on TikTok debunk harmful myths with a shared five-step process that pairs each science-based rebuttal directly with the original misinformation video.","keywords":["misinformation","debunking","TikTok","health professionals","content moderation","duet","stitch","mental health"],"falsifier":"Audit a random sample of credentialed health professionals' TikTok accounts in nutrition and mental health without filtering by #duet or #stitch hashtags; if a substantial fraction of their debunking videos are standalone (no stitch, no duet, no tagging of the original creator), then the claimed universal response stage and the 'symmetric content-to-content' feature would be contradicted.","tokens_in":31835,"feed_emoji":"🩺","tokens_out":6387,"duration_ms":60574,"temperature":0.7,"pith_summary":"The paper argues that when platforms fail to directly refute health misinformation, credentialed health professionals on TikTok mount their own counter-influence: they pick a trending false video, craft a science-based rebuttal, and attach it to the original as a duet or stitch. Analyzing open-ended surveys from 14 such professionals working on nutrition and mental health, the authors identify a common five-stage process — initiation, selection, creation, response, and post-debunking management — with three defining features: targets are trending claims seen as directly harmful, the rebuttal is a video-to-video response of the same format and magnitude as the original claim, and the content is grounded in scientific evidence and the debunker's clinical experience rather than third-party fact-checking services. If this model is right, it gives researchers and platforms a first formal description of a grassroots, professional-led alternative to warning labels and community notes, and a basis for testing whether such efforts actually change beliefs. The paper deliberately does not measure the effectiveness of these debunking videos on audiences, framing that as future work.","feed_headline":"Health pros debunk TikTok myths with one five-step playbook","feed_subtitle":"A survey of 14 professionals maps how they pick viral false videos, stitch a science-based reply, and manage the backlash.","key_machinery":"The central object is the five-stage 'Debunk-It-Yourself' process model (initiation, selection, creation, response, and post-debunking management), built from inductive thematic coding of survey responses with a Cohen's kappa of 0.8334 and backed by metadata for 1,649 videos from the 14 participants. The mechanism that carries the argument is the symmetric video-to-video response made possible by TikTok's stitch and duet affordances: the debunker's video plays alongside the original, so the correction arrives in the same format, on the same feed, and with the same potential virality as the misinformation it counters.","core_discovery":"In the paper's own terms, the discovery is that 'Debunk-It-Yourself' is not a scattering of individual reactions but a shared, repeatable process. The professionals do not actively search for misinformation; the platform's recommendation feed surfaces it to their For You Page, followers tag them, or patients, students, friends, and family ask them about a claim. They then select targets by virality, blatant falsity, and visible audience susceptibility (such as comments saying 'I didn't know this'), craft a rebuttal based on academic evidence, clinical experience, and peer consultation — explicitly avoiding fact-checking services — and respond in kind by stitching or duetting the original video, tagging its creator, or reusing its hashtags. After posting, they decide whether to report the video, monitor comments for contradictions, and appeal when TikTok wrongly labels their own debunking content as misinformation. The paper presents this as the first documentation of a symmetric, content-to-content countermeasure that follows the Debunking Handbook's prescription more literally than platform moderation does.","pith_inferences":["Editorial inference: the 'symmetric response' feature is platform-dependent. On platforms without stitch or duet affordances, the same professionals would likely fall back to comments or standalone videos, so this defining feature may describe TikTok's ecology rather than a stable property of the debunkers themselves.","Editorial inference: the paper does not test whether DIY debunking changes viewers' beliefs; a natural next experiment would compare belief correction from stitched rebuttals versus warning labels using the same misinformation videos as stimuli.","Editorial inference: the recruitment pipeline (hashtags like #adhd, #anorexia, #duet, and #stitch) may under-sample debunkers who respond in other formats, so a hashtag-free audit of credentialed health accounts would reveal how universal the five stages really are.","Editorial inference: the duet's traffic-boosting effect cuts both ways — if the debunking video sends viewers to the original, the net effect on misinformation exposure could be positive, a tension the participants acknowledged but the study did not quantify."],"forward_implications":["The five-stage process can serve as a template for studying or building user-led debunking on other short-video platforms, such as Instagram Reels or YouTube Shorts, that offer similar pairing features.","Platforms could prioritize evidence-based debunking content in recommendations, prebunk before harmful claims spread, compensate debunkers, and stop mislabeling debunking videos — the concrete steps the participants themselves recommended.","The model exposes structural tensions: debunkers fight an algorithm that favors sensational content, risk shadowbanning and harassment, and may inadvertently send traffic to the original misinformation video when they duet it.","A rogue or biased debunker is a real threat in this paradigm, since individual debunkers are not accountable to any professional body or platform policy; the paper suggests professional medical associations could impose standards.","Because the study excluded political content and did not measure audience effects, the model's generality to other misinformation domains and its actual counter-influence effectiveness remain open questions that the paper explicitly flags as future work."],"supporting_citations":[{"why":"Supplies the four-step debunking prescription (state truth, point to misinformation, explain why wrong, restate truth) that the paper claims DIY materializes in video form.","marker":"[44]"},{"why":"Documents TikTok's hashtag- and keyword-based warning-label moderation, the platform intervention that DIY is contrasted against throughout the paper.","marker":"[45]"},{"why":"Describes community-note-style crowdsourced moderation, the user-led alternative that DIY is compared with as a symmetric and expert-driven response.","marker":"[27]"},{"why":"Reports health professionals actively pushing back on TikTok health myths, the phenomenon this study formally maps.","marker":"[67]"},{"why":"Documents the stitch and duet affordances that enable the symmetric video-to-video debunking response.","marker":"[78]"},{"why":"Provides prior evidence that health professionals correct misinformation through comments and replies, the asymmetric baseline from which DIY departs.","marker":"[9]"},{"why":"Supplies the thematic-analysis guidelines used to code the survey data and build the five-stage process model.","marker":"[14]"},{"why":"Characterizes TikTok's visibility moderation and algorithm-driven viewing, the environment that shapes how debunkers encounter and respond to misinformation.","marker":"[93]"},{"why":"Previous work on the effectiveness of TikTok debunking videos, which motivates the paper's focus on the debunkers' own process and supply side.","marker":"[12]"}],"fun_headline_variants":["DIY debunking: health pros' 5-step TikTok counterattack","Health pros stitch science onto TikTok myths in 5 steps","From viral claim to science reply: the 5-step DIY debunking","Health pros' DIY playbook: 5 steps to debunk TikTok misinformation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole five-stage model rests on the assumption that the 14 survey respondents, recruited through a hashtag-based TikTok search and manual credential vetting, accurately represent the full population of health-professional debunkers on nutrition and mental health; any debunkers who never use those hashtags or never stitch or duet are invisible to the study.","fun_headline_variants_meta":{"raw":{"variants":["DIY debunking: health pros' 5-step TikTok counterattack","Health pros stitch science onto TikTok myths in 5 steps","From viral claim to science reply: the 5-step DIY debunking","Health pros' DIY playbook: 5 steps to debunk TikTok misinformation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001021,"raw_usage":{"total_tokens":4371,"prompt_tokens":1072,"completion_tokens":3299,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":688,"completion_tokens_details":{"reasoning_tokens":3221}},"tokens_in":688,"tokens_out":3299,"duration_ms":24786,"temperature":1.0,"reasoning_tokens":3221,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T20:59:14.229019+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Audit a random sample of credentialed health professionals' TikTok accounts in nutrition and mental health without filtering by #duet or #stitch hashtags; if a substantial fraction of their debunking videos are standalone (no stitch, no duet, no tagging of the original creator), then the claimed universal response stage and the 'symmetric content-to-content' feature would be contradicted.","supporting_citations":[{"cited_title":"Kendou, D","cited_arxiv_id":null,"evidence_quote":"Supplies the four-step debunking prescription (state truth, point to misinformation, explain why wrong, restate truth) that the paper claims DIY materializes in video form."},{"cited_title":"Learn the Facts about COVID-19","cited_arxiv_id":null,"evidence_quote":"Documents TikTok's hashtag- and keyword-based warning-label moderation, the platform intervention that DIY is contrasted against throughout the paper."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports health professionals actively pushing back on TikTok health myths, the phenomenon this study formally maps."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the stitch and duet affordances that enable the symmetric video-to-video debunking response."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides prior evidence that health professionals correct misinformation through comments and replies, the asymmetric baseline from which DIY departs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the thematic-analysis guidelines used to code the survey data and build the five-stage process model."},{"cited_title":"misinformation","cited_arxiv_id":null,"evidence_quote":"Characterizes TikTok's visibility moderation and algorithm-driven viewing, the environment that shapes how debunkers encounter and respond to misinformation."}],"review_version":1}