{"id":"c482315c-720d-4bef-b1de-def026f98afb","arxiv_id":"2501.18210","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Women on Xiaohongshu re-appropriate baby-food hashtags as an audience-blocking tactic, but the strategy degrades as it becomes popular.","lead":"This study analyzed 5,800 posts and 24 interviews on Xiaohongshu to show how women re-purpose hashtags like #BabySupplementalFood in unrelated posts to try to keep their content away from male audiences. It documents the tactic's rise, why women use it, and why it stops working as it goes viral.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper documents hashtag-content mismatch, but the claim that #BSF 'blocks male users' rests on untested folk theory; no outcome data links the hashtag to reduced male viewership, and P12's report directly contradicts it.","rationale":"The reader's weakest assumption identifies exactly the point on which the central claim hinges: the paper treats hashtag-topic relevance as a proven lever for excluding male audiences, but the only evidence is users' folk theories and two self-reports, with a documented counterexample. My read agrees and adds specificity: Section 5.1's causal phrasing overreaches the quantitative result. The 84.8% irrelevant-post figure proves that #BSF has been detached from its literal meaning, and topic modeling shows diverse content, but this does not demonstrate a reduction in male viewership. The interviews establish intent and belief, not effect. The failure case of P12 is especially important because it is not noise: a post using #BSF received overwhelmingly male attention, suggesting the mechanism is unreliable or context-dependent. The paper also explicitly acknowledges later that #BSF became less effective as it gained popularity (Sections 5.1.3 and 5.3.4), which further undercuts the unqualified blocking claim. This is an internal overreach rather than a disagreement with external consensus, so it is a correctness risk in the paper's own terms. Despite this, the descriptive contribution—documenting a communal, deliberate hashtag re-appropriation practice and its evolution—is well supported by convergent quantitative and qualitative evidence, including relevance analysis, topic modeling, co-occurrence networks, and 24 interviews. Therefore the appropriate verdict remains CONDITIONAL: the phenomenon is real, but the headline audience-control effect needs independent validation before it can be stated as demonstrated.","tokens_in":34828,"tokens_out":3277,"duration_ms":37516,"concrete_test":"Conduct a matched-pair field audit: create fresh Xiaohongshu accounts with identical neutral profiles; publish the same content with and without #BSF at matched times; and infer viewer gender from public engagement cues (profile metadata of commenters and likers). If #BSF-tagged posts do not show a significantly lower male-engagement ratio than untagged control posts across repeated trials, the 'blocking male users' mechanism is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the inference in Section 5.1 that the 84.8% irrelevant-post rate 'demonstrat[es] its effect in blocking male users.' The relevance analysis in Section 4.2.2 only establishes hashtag-content mismatch; it says nothing about who actually saw the posts. The causal mechanism—that Xiaohongshu's recommender uses hashtag-topic relevance to avoid recommending baby-food-tagged posts to men—is folk theory, supported only by P05 and P08's self-reports in Section 5.1.2 and contradicted by P12's report of a #BSF post receiving over 20,000 primarily male views in Section 5.2.2. P14's account shows the post was recommended to a lesbian woman, which is compatible with interest-based matching but not with reliable male exclusion. Since the paper's central contribution is audience control, not merely symbolic signaling, the claim requires outcome data on audience composition or algorithmic exposure. Without such data, the re-appropriation practice is well documented, but its audience-blocking effect remains unverified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper investigates the re-appropriation of the #Baby Supplemental Food (#BSF) hashtag on the recommendation-driven platform Xiaohongshu, where women users deliberately attach a baby-food hashtag to posts unrelated to baby food in the hope of keeping male users out of their audiences. The authors combine a quantitative analysis of 5,800 scraped posts (hashtag–post relevance classification, BERTopic modeling, LIWC expression analysis, and hashtag co-occurrence networks) with 24 semi-structured interviews of Xiaohongshu users. They report that 84.8% of #BSF posts are unrelated to baby food, describe the main topics within those posts, document the emergence of derived hashtags (#BSF(TC), #Male Sterilization, #185 Handsome Guy), and identify motivations including platform strategy shifts, gender-based harassment, and ineffective reporting/blocking tools. The paper frames the practice as a form of everyday digital feminist resistance and proposes a design concept called 'dynamic audience control.'","tokens_in":35018,"tokens_out":3481,"duration_ms":35236,"significance":"If the central claim is accepted, the paper makes a useful contribution to HCI/CSCW scholarship on algorithmic folk theories, feminist HCI, and user self-governance on recommendation-driven platforms. Its strengths include a genuinely mixed-methods design, a transparent positionality statement, a detailed codebook and interview protocol in the appendices, and a rich set of interview excerpts that capture disagreement and nuance within the user community. The documentation of hashtag re-appropriation as a communal, evolving tactic, including the negative case reported by P12, is valuable empirical material. However, the paper's headline interpretive claim—that #BSF actually blocks male audiences—is not established by the data presented; the quantitative evidence shows hashtag–content mismatch, not audience composition, and the interview evidence is mixed. The contribution is therefore better described as documenting an intended and believed audience-management tactic with uncertain effectiveness, rather than demonstrating a working method of audience control.","major_comments":[{"comment":"The inference from the 84.8% irrelevance rate to 'demonstrating its effect in blocking male users' is not supported by the data. Hashtag–post relevance is a property of post text, while the blocking claim is a claim about the recommender's audience distribution. The quantitative analysis establishes that #BSF posts are mostly not about baby food, but it does not establish who saw those posts or that male viewership was reduced. The interview support for effectiveness rests on two self-reports (P05, P08), while P12's account in Section 5.1.2 and Section 5.2.2 reports a #BSF post receiving more than 20,000 views primarily from men, which directly contradicts a reliable blocking effect. Please either present outcome data on audience composition or algorithmic exposure, or reframe the claim as 'users intend and believe this tactic blocks male audiences, with mixed and largely unverified effectiveness.'","section":""},{"comment":"The headline quantitative result rests on a GPT-4o-mini classifier validated on 100 manually annotated post titles, with an overall accuracy of 94% on a test set containing 32 relevant and 68 irrelevant posts. No confidence interval, per-class precision/recall, or agreement statistics for the final classification are reported. Since the 84.8% irrelevance estimate is load-bearing for the entire quantitative narrative, the paper should report a binomial confidence interval for that prevalence, class-level performance metrics, and ideally a larger or randomly sampled validation set. Without this, readers cannot assess how much of the 84.8% figure could be classifier noise or bias.","section":""},{"comment":"The authors acknowledge in Section 6.5 that it is 'infeasible to determine if the data is biased or fully representative' for posts scraped from a recommendation-driven platform. Given this, the quantitative claims in Section 5.1.1 (e.g., 84.8% irrelevant, topic cluster sizes) should be explicitly presented as descriptive statistics for the collected sample rather than as population estimates. The current wording in the abstract and findings, which presents these figures without such hedging, overstates their scope.","section":""},{"comment":"The paper's own evidence shows that the blocking effect was contested and degraded over time: P12 reported a spectacular failure, P10 and P06 described reduced effectiveness, and Section 5.3.4 documents traffic-driven co-optation. Yet the abstract and conclusion state that the practice 'blocks male audiences' without qualification. The Discussion already contains the right nuance about hashtag adaptability and limitations, so the central claims in the abstract and findings should be aligned with that nuance.","section":""}],"minor_comments":[{"comment":"The text says 'Full results of the post-expression analysis can be found in Appendix 8,' but the figure is numbered as Figure 8 in Appendix E; please correct the cross-reference. The caption of Figure 8 mentions an upward arrow (↑) indicating direction of difference, but no arrows appear in the figure itself; either add the arrows or remove the mention from the caption.","section":"Section 5.1.3 and Figure 8"},{"comment":"Several topic keywords in Table 4 appear to be artifacts or unexplained tokens (e.g., '16592' in topic 19, '6025' in topic 20, and 'ahhhh' in topic 19). Please clean or explain these tokens, or note that they are raw c-TF-IDF outputs.","section":"Table 4"},{"comment":"The participant recruitment mix is described clearly, but the paper does not state whether the interviewees who reported using #BSF were users of the derived hashtags as well; adding a column or a sentence on which participants used which hashtags would strengthen the connection between the interview data and the quantitative hashtag analysis.","section":"Section 4.3.1 / Table 2"},{"comment":"Reference [57] is a Chinese-language source with a translated title; please provide the original title or a note that it is in Chinese, for consistency with the other Chinese-language citations.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a good fit for a CHI/CSCW audience and the qualitative material is rich. My main concern is not with the existence of the practice—the documentation is convincing—but with the strength of the causal claim about blocking effectiveness. The authors should be given the opportunity to either add audience-outcome evidence or resubmit with the claim appropriately hedged. The classifier validation issue is secondary but should also be addressed, since the 84.8% figure is cited prominently."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, this one is worth a look. The paper documents a real, novel phenomenon: Chinese women on Xiaohongshu re-appropriating #BabySupplementalFood to keep male users out of their audiences, and it offers a process model (blocking, attracting, evolving) plus a useful two-axis framing (attract/block × relevant/irrelevant). The mixed-methods design is solid: relevance classification, topic modeling, LIWC expression differences, co-occurrence networks, and 24 interviews triangulate on the practice and its spread. They even include P12’s case where the tactic failed badly, which is honest and strengthens the credibility of the qualitative analysis.\n\nThe main soft spot is the load-bearing inference in Section 5.1 that the 84.8% irrelevant-post rate demonstrates the hashtag’s effect in blocking male users. It does no such thing on its own. Relevance mismatch shows re-appropriation, not audience composition. The blocking effect rests on two interviewees’ impressions (P05, P08), one participant’s counterexample (P12), and an untested folk theory about Xiaohongshu’s recommender using hashtag-topic relevance to avoid pushing baby-food posts to men. The paper would be on firmer ground if it framed the contribution as user intention and perceived effect, and noted the platform mechanism as unverified. That is a framing fix, not a fatal flaw, because the core descriptive finding—that women communally re-appropriate hashtags as an audience-management tactic—is well supported by convergent evidence.\n\nMinor issues: the classifier (GPT-4o-mini, 94% accuracy on a 100-post manual label set) is fine for a first pass but deserves confidence intervals and a larger validation set; scraping representativeness is acknowledged as unknowable; and no data or code are released, which limits reproducibility.\n\nThis is a publishable case study that will be useful to anyone working on algorithmic folk theories, platform governance, or feminist HCI. It deserves a serious referee; the weak causal claim is easy to fix with rephrasing and a limitations paragraph. I would accept for review, and I would cite the empirical case.","headline":"A well-evidenced case study of hashtag re-appropriation as an audience-control tactic; the causal blocking claim should be softened to perceived effect.","tokens_in":35582,"tokens_out":2093,"would_cite":true,"duration_ms":19974,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A baby-food hashtag has become a gender filter on Xiaohongshu.","keywords":["hashtag re-appropriation","recommendation algorithms","audience control","Xiaohongshu","safe spaces online","gender-based harassment","folk theories","digital feminism"],"falsifier":"A controlled field experiment would settle it: publish identical posts on Xiaohongshu with and without #BSF using fresh accounts, and compare the gender composition of viewers and engagement; if the tag does not shift the audience's gender composition, the audience-control effect is not real.","tokens_in":34606,"feed_emoji":"🏷️","tokens_out":8541,"duration_ms":76418,"temperature":0.7,"pith_summary":"This paper argues that women on Xiaohongshu, a recommendation-driven social platform, have turned hashtags into audience filters. By attaching '#Baby Supplemental Food' to posts about makeup, pets, relationships, and daily life, they try to stop male users from being shown their content. The study combines analysis of 5,800 scraped posts with interviews of 24 users to trace how the practice began, spread, and changed. If the authors are right, users can reclaim partial control over algorithmic distribution through collective semantic misdirection, turning recommendation systems into a site of everyday resistance.","feed_headline":"Baby-food hashtag becomes a gender filter on Xiaohongshu","feed_subtitle":"Study of 5,800 posts and 24 interviews traces how a baby-food tag filters audiences and fades.","key_machinery":"The central object is the re-appropriated hashtag used as a semantic misdirection device. #BSF ('Baby Supplemental Food') is a topical label that the target audience, men, is presumed to find uninteresting, and the recommendation algorithm uses hashtags as a distribution signal, so the tag acts as an algorithmic audience filter rather than a content descriptor. The paper also organizes the practice on two axes, audience management (attract vs. block) and content relevance (relevant vs. irrelevant), yielding three hashtag uses: regular usage, traffic-driving, and guarding against male users. The feedback loop that carries the argument is that blocking success attracts more users, viral popularity brings traffic-driven co-optation, effectiveness drops, and women derive new hashtags such as #BSF(TC), #HG, and #MS to restart the loop.","core_discovery":"The central claim is that the re-appropriation of #BSF is deliberate, communal, and effective-enough audience control, not random misuse or a simple meme. Of 3,311 posts using #BSF, 84.8% were unrelated to baby food and instead covered food, outfits, love, makeup, weight loss, parenting tips, packaging, and cats. Interviews with 24 users show that women learned the trick from peers, discussed it in private chat groups, and believed it reduced male viewership and harassment. The paper describes the process as 'blocking, attracting, and evolving': the hashtag originally blocked men, then attracted women seeking a safe space, then attracted advertisers and thirst traps, and finally lost power, leading women to create derived hashtags such as #BSF(TC), #Male Sterilization, and #185 Handsome Guy. The authors read this as everyday resistance within digital feminism and use it to argue that recommendation-driven platforms need 'dynamic audience control' features.","pith_inferences":["We infer that the same semantic-misdirection trick should work on other recommendation platforms: any group can pick a topic presumed boring to an out-group and use its hashtag as an audience filter, so analogous tags are likely to appear on TikTok, Instagram, and Douyin.","The paper's own failure case (P12) suggests the tactic is brittle: for content the algorithm is likely to push hard anyway, such as bikini photos, the hashtag signal can be overwhelmed, so effectiveness likely depends on post type and algorithm volatility.","A testable extension follows directly: a repeated scraping study could monitor hashtag-topic divergence over time and predict when a blocking hashtag is about to lose its power, before users abandon it."],"forward_implications":["Women can partially exclude unwanted male audiences from recommendation-driven feeds without platform support, by exploiting the algorithm's reliance on hashtag topic signals.","The tactic is self-limiting: as a blocking hashtag goes viral, traffic-driven users and men adopt it, and the blocking effect erodes, forcing women to invent derived hashtags.","Existing safety tools, including reporting, blocking, and privacy settings, do not meet creators' need to control who sees a post, which is why users turn to algorithmic workarounds.","Hashtag use can be described on two dimensions, audience management (attract vs. block) and content relevance (relevant vs. irrelevant), which distinguishes regular usage, traffic-driving, and guarding against male users.","The practice constitutes a form of everyday resistance that shifts some control over content distribution from the platform to users, even if users do not frame it as activism."],"supporting_citations":[{"why":"Supplies the folk-theory formation cycle that explains how women inferred the hashtag's blocking effect and shared it.","marker":"[23]"},{"why":"Defines folk theories as intuitive theories guiding user behavior toward algorithmic systems, the lens through which the paper interprets users' beliefs about hashtag effectiveness.","marker":"[24]"},{"why":"Shows that algorithms suppress marginalized identities and that users engage in algorithmic resistance, grounding the interpretation of hashtag blocking.","marker":"[51]"},{"why":"Establishes hashtags as infrastructure for safe spaces in Black Twitter, giving the comparative case for audience control and context collapse.","marker":"[54]"},{"why":"Provides the concept of narrative agency in hashtag activism that frames #BSF posts as collective storytelling with audience-shaping power.","marker":"[100]"},{"why":"Documents how marginalized communities develop new lexical hashtag variants as old ones are co-opted or moderated, supporting the paper's evolution analysis.","marker":"[14]"},{"why":"Provides the comparable TikTok tactic of starting with unappealing topics to filter out male viewers, showing the strategy is not platform-specific.","marker":"[87]"},{"why":"Media report that first described #BSF as a wall against men on Xiaohongshu, motivating the study.","marker":"[73]"},{"why":"Supplies the everyday-resistance framework used to interpret the hashtag work as semi-conscious resistance rather than overt activism.","marker":"[89]"}],"fun_headline_variants":["Women hijack baby-food tag to block men on Xiaohongshu","Xiaohongshu women weaponize #BabyFood to block men","Baby-food hashtag is women's secret weapon on Xiaohongshu","How #BabyFood became a password for women on Xiaohongshu"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim that #BSF actually blocks male users depends on the unverified assumption that Xiaohongshu's recommendation algorithm uses hashtag topic as a strong signal for who gets shown a post; the paper's support is folk theory plus two interviewees' impressions, and one interviewee reports the hashtag failing completely.","fun_headline_variants_meta":{"raw":{"variants":["Women hijack baby-food tag to block men on Xiaohongshu","Xiaohongshu women weaponize #BabyFood to block men","Baby-food hashtag is women's secret weapon on Xiaohongshu","How #BabyFood became a password for women on Xiaohongshu"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001396,"raw_usage":{"total_tokens":5635,"prompt_tokens":922,"completion_tokens":4713,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":4630}},"tokens_in":538,"tokens_out":4713,"duration_ms":35489,"temperature":1.0,"reasoning_tokens":4630,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T00:16:44.506962+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A controlled field experiment would settle it: publish identical posts on Xiaohongshu with and without #BSF using fresh accounts, and compare the gender composition of viewers and engagement; if the tag does not shift the audience's gender composition, the audience-control effect is not real.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows that algorithms suppress marginalized identities and that users engage in algorithmic resistance, grounding the interpretation of hashtag blocking."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes hashtags as infrastructure for safe spaces in Black Twitter, giving the comparative case for audience control and context collapse."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the concept of narrative agency in hashtag activism that frames #BSF posts as collective storytelling with audience-shaping power."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the comparable TikTok tactic of starting with unappealing topics to filter out male viewers, showing the strategy is not platform-specific."},{"cited_title":"Wall Against Men","cited_arxiv_id":null,"evidence_quote":"Media report that first described #BSF as a wall against men on Xiaohongshu, motivating the study."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the everyday-resistance framework used to interpret the hashtag work as semi-conscious resistance rather than overt activism."}],"review_version":1}