{"id":"e6d89f98-0d96-4f41-a5ad-c8e6b4fce598","arxiv_id":"2504.20308","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A systematic review of 66 studies shows youth online safety research in the Global South is concentrated in Asia, dominated by quantitative surveys, and centered on cyberbullying rather than marginalized populations or policy.","lead":"This paper systematically reviews 66 studies on youth online safety in the Global South and maps where, with whom, and how that research was done. It finds the field is Asia-heavy, survey-heavy, and focused on cyberbullying, and it argues for culturally situated, youth-centered methods.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline landscape claims rest on a corpus that may not represent Global South research: three Western-index databases, English/Spanish-only searches, and single-screener selection could manufacture the Asia-heavy, marginalized-youth-absent pattern.","rationale":"The most load-bearing assumption is that the constructed corpus is representative of the relevant Global South literature. If the retrieval frame excludes Portuguese-, French-, Arabic-, and Chinese-language research and regional indexing systems, then the observed regional imbalance and the apparent absence of marginalized-youth work are not properties of the field but artifacts of the search strategy. This directly threatens the strongest claim about concentrated Asian focus and limited marginalized-youth research. I agree with the reader's weakest assumption; the count inconsistencies are important but secondary, because even corrected counts would likely still show a quantitative, Asia-heavy, cyberbullying-focused literature. A replication across additional databases and languages, with dual screening, would settle the question. Since the paper is a systematic review whose central value is descriptive accuracy, the appropriate posture remains conditional: the contribution is useful and directionally plausible, but it should not be treated as a definitive map until the corpus is shown to be robust to the missing literatures or the claims are explicitly narrowed to the indexed English/Spanish corpus.","tokens_in":31462,"tokens_out":7718,"duration_ms":80239,"concrete_test":"Re-run the PRISMA search adding Scopus, IEEE Xplore, SciELO, Redalyc, AJOL, and CNKI, with the full query translated into Portuguese, French, Spanish, Arabic, and Chinese, and with two independent screeners applying the same inclusion criteria to title and abstract; then recompute Tables 3 and 4. If the Asia share falls below half or the Africa/South America share materially increases, the regional-concentration headline is an artifact of the original search frame. As a second arm of the same replication, run a targeted query for marginalized youth using terms such as 'LGBTQ', 'disability', and 'autism' paired with Global South country names, without requiring the 'online safety' keyword, to test whether the apparent absence of this literature is an artifact of search vocabulary.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that Global South youth online safety research is Asia-concentrated, quantitative-dominant, cyberbullying-focused, and mostly ignores marginalized youth—depends entirely on the corpus built in Sections 4.1 and 4.2. The corpus is retrieved from ACM Digital Library, Web of Science, and PubMed, with queries in English and Spanish, and title/abstract screening done by the first author alone. These choices are not neutral. Most Latin American HCI and health literature is in Portuguese and indexed in SciELO/Redalyc; Francophone African and Middle Eastern work appears in French/Arabic and regional journals; Chinese studies are indexed in CNKI. A study about LGBTQIA+ or disabled youth that frames the issue as harassment or well-being rather than 'online safety' would not be retrieved. The observed outcomes—fewer African and South American studies, only two marginalized-youth papers, cyberbullying as the dominant risk—are exactly what this retrieval frame would produce even if the underlying field were more balanced. Section 6.5 admits that only four Spanish-language papers were found, but does not quantify how many relevant non-English or non-indexed records were missed, so the bias is unmeasured. Internal count inconsistencies (Section 5.3 says quantitative N=66 while Table 4 says N=46; Section 5.5 gives 44 plus 20 risk studies totaling 64, not 66) further reduce confidence that the coded frequencies can support the headline.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a systematic review of 66 studies published between 2014 and 2024 on youth online safety in the Global South, following a PRISMA-inspired process and coding the corpus along four dimensions: where the research was conducted, who was studied, how it was conducted, and what risks and safety strategies were addressed. The authors report that the literature is concentrated in Asian countries, dominated by quantitative survey methods, and focused primarily on cyberbullying, with limited attention to marginalized youth populations and a strong emphasis on cultural and social factors. The paper concludes with methodological recommendations and a future research agenda calling for qualitative, participatory, and culturally situated approaches.","tokens_in":31724,"tokens_out":4581,"duration_ms":46640,"significance":"If the descriptive landscape claims are reliable, this is a valuable contribution to HCI and CSCW: it extends the earlier review by Pinter et al. (2017), provides a useful codebook for the field, and explicitly addresses positionality, which is still uncommon in systematic reviews. The paper gives due credit to regional scholars and identifies concrete gaps, such as the near-absence of LGBTQIA+ and disability-focused studies. However, the reliability of the headline claims is weakened by internal count inconsistencies and by a retrieval and screening strategy that may not represent the full Global South literature, so the current manuscript needs revision before its central empirical claims can be fully trusted.","major_comments":[{"comment":"The count of quantitative studies is internally inconsistent. The text states 'a significant predominance of articles using quantitative methods (N=66)', but Table 4 reports Quantitative (46), Qualitative (17), and Mixed-method (3), which sum to 66. The surrounding sentences also report qualitative N=17 and mixed-method N=3, so the N=66 for quantitative methods appears to be a typographical error. Since 'predominance of quantitative methods' is one of the three headline findings in the abstract, this discrepancy must be corrected and all counts reconciled with Table 4.","section":"Section 5.3 and Table 4"},{"comment":"Several other frequency counts do not add up. In Section 5.5, the text says 'Most studies (N=44) investigated risks among five major categories' and 'The remaining 20 studies' used a broad definition, which totals 64, not 66; yet the enumerated category counts for cyberbullying (29), privacy (14), sexual risks (8), inappropriate content (2), and technology addiction (3) sum to 56, not 44 or 64. In Section 5.1, the text reports Africa N=13, while Table 4 lists Africa (12), and the regional counts for Asia (48), Africa, and South America (8) together exceed 66. These inconsistencies directly affect the claim that cyberbullying is the most prominent risk and the claim about regional concentration, so a full audit of the coded frequencies is required.","section":"Sections 5.1, 5.5, and Table 4"},{"comment":"The sampling frame is too narrow to support the unqualified landscape claims in the abstract. The search is limited to ACM Digital Library, Web of Science, and PubMed, with queries in English and Spanish, and the initial title/abstract screening was performed by the first author alone. This design likely underrepresents Portuguese-language literature indexed in SciELO and Redalyc, Francophone African and Middle Eastern research, and Chinese-language work in CNKI, all of which are relevant to the Global South. The observed pattern of few African/South American studies and few marginalized-youth studies could be an artifact of this retrieval frame rather than a true property of the field. Section 6.5 acknowledges that only four Spanish-language papers were found, but it does not quantify how many non-English or non-indexed records were missed. The authors should either broaden the search and screening process or explicitly reframe the conclusions as describing the indexed English/Spanish literature in these three databases, rather than 'the Global South' as a whole.","section":"Sections 4.1, 4.2, and 6.5"}],"minor_comments":[{"comment":"The text reports 'survey-based studies (N=43)' while Table 4 lists 'Survey (52)' as the method count; please clarify whether studies could be assigned multiple methods and reconcile the numbers.","section":"Section 5.3 and Table 4"},{"comment":"References [54] and [55] appear to be the same work (Hendricks et al., 2020, 'Cyber-Harassment Victimization Among South African LGBTQIA+ Youth'), which creates confusion in Section 5.2 where the paper says only two marginalized-population studies were found and then appears to cite the same study twice.","section":"References [54] and [55]"},{"comment":"There is a typographical error in the phrase 'youth' consider online risks and safety to be on a spectrum'; the apostrophe should be removed.","section":"Section 3.1"},{"comment":"The footnotes say that both BrazilIHC and AfriCHI proceedings have been published in ACM 'since the seventeenth edition of the conference held in 2016', which is not accurate for either conference; please verify the edition numbers and years.","section":"Section 4.1, footnotes 3-5"},{"comment":"The 'Authors' affiliation' row lists countries without counts or percentages; adding per-country frequencies would improve transparency and align this row with the other coded dimensions.","section":"Table 4"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses an important and under-served topic, and the authors' positionality statement is a strength. However, the internal inconsistencies in the coded counts are not merely cosmetic: they affect the paper's three headline claims. The corpus representativeness concern is also significant; I would encourage the editor to ask the authors to either strengthen the search or soften the claims to match the actual retrieval frame. These issues are fixable within the manuscript's scope, so I do not recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is the first systematic review of youth online safety research in the Global South, coding 66 papers from 2014–2024. That alone gives it value. The headline findings are plausible: Asia dominates, surveys dominate, cyberbullying is the most studied risk, and culture is the most discussed social factor. The authors are transparent about positionality and limitations, and the appendix codebook is a helpful resource.\n\nThe soft spots are real and not cosmetic. Section 5.3 reports quantitative methods as N=66 when it clearly means 46, and Table 4 lists 46 quantitative, 17 qualitative, 3 mixed-method, which sums to 66 — so the text contradicts its own table. Section 5.5 reports 44 plus 20 risk studies, which sums to 64, not 66, and the five risk categories in Table 4 sum to 56, not 44. There are similar inconsistencies with survey counts (text says 43, table says 52) and Asia totals (text adds to 46, table says 48). For a systematic review, those numbers need a careful audit before the specific frequencies can be trusted.\n\nThe deeper issue, which the stress-test note correctly flags, is the sampling frame. Three databases — ACM DL, Web of Science, PubMed — searched in English and Spanish systematically undercounts Portuguese-language Brazilian HCI, French and Arabic work, and anything indexed only in regional venues. The authors admit only four Spanish-language papers made it and no Caribbean or Oceania representation, but they never quantify the potential miss. So the “Asia-heavy, Africa/Latin America-light” landscape might be partly an artifact of the search. That is the biggest editorial risk: the abstract states these patterns as facts about the field when they are really properties of the reviewed corpus.\n\nThe single-screener title/abstract phase is a minor weakness, since the full-text review was done by two people and the coding appears consensus-based. Not a dealbreaker.\n\nBottom line: this is a useful map with a shaky legend. The synthesis is needed and the core direction is almost certainly right, but the authors need to fix the count inconsistencies, reconcile text and tables, and frame the geographic and methodological patterns as characteristics of the reviewed corpus rather than of Global South literature as a whole. I would send it to peer review — it deserves serious referee time — and I would cite it after those fixes are made.","headline":"Useful first synthesis of Global South youth online safety research, but the count inconsistencies and unmeasured retrieval bias mean the findings should be treated as suggestive until fixed.","tokens_in":32299,"tokens_out":2700,"would_cite":true,"duration_ms":28086,"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":"A systematic review of 66 studies finds that Global South youth online safety research is Asia-heavy, survey-dominated, and centered on cyberbullying, with culture the most emphasized social factor.","keywords":["youth online safety","Global South","systematic review","cyberbullying","culture","marginalized youth","quantitative surveys","education and awareness"],"falsifier":"Re-run the review's Boolean query in regional and non-English publication venues that were not among the three selected databases, and code the additional hits with the same where/who/how/what scheme; if the added set contains many qualitative, policy-focused, or Latin American and African studies, the claims about regional concentration, methodological dominance, and the absence of marginalized-youth research would fail as statements about the field. A second check is to re-code all 66 full texts for whether any demographic or risk data about LGBTQIA+ or disabled youth appear even when not declared in the title or abstract.","tokens_in":31272,"feed_emoji":"🛡️","tokens_out":10302,"duration_ms":101952,"temperature":0.7,"pith_summary":"This paper sets out to establish that the research base on youth online safety in the Global South is not yet a balanced, globally inclusive field. Reviewing 66 studies published between 2014 and 2024, the authors find the literature concentrates in Asian countries, leans heavily on one-time quantitative surveys, gives little attention to LGBTQIA+ and disabled youth, and treats cyberbullying as the dominant risk. They argue that cultural factors—family structures, collectivist norms, religious values, gender expectations—are the most influential social forces shaping both how risks appear and how safety strategies are designed in these regions. If the review is right, future work should replace imported one-size-fits-all interventions with situated, participatory, and culturally responsive methods led with youth in the Global South.","feed_headline":"Youth online safety research skews Asian and survey-based","feed_subtitle":"A 66-paper review finds cyberbullying dominates and marginalized youth are nearly absent.","key_machinery":"The carrying object is the coded corpus of 66 papers plus the qualitative scheme applied to it. The corpus was assembled through a systematic protocol: Boolean queries combining online-safety, youth, and Global South terms (including a full list of countries), restricted to 2014–2024 and to English- and Spanish-language publications, with inclusion criteria requiring peer review, a focus on youth aged 19 or younger, direct treatment of online risk, and a contribution to HCI, followed by title/abstract and full-text screening. The analytic machinery is a four-dimensional codebook—where the research took place, who was studied, how data were collected and what kind of contribution resulted, and what risks, social factors, and safety strategies were examined—and applying it consistently is what converts the article list into claims about regional concentration, methodological dominance, and cultural emphasis.","core_discovery":"The central discovery, stated as the authors would state it, is a description of what a decade of research has and has not covered. Across the 66 papers, the settings cluster in Asia, with China (11), India (9), Malaysia (6), and South Africa (4) appearing most often; surveys are the dominant data-collection method (43 of 66), qualitative studies number 17, mixed-method studies 3, and longitudinal designs 1. Cyberbullying and online victimization is the most studied risk category (29 studies), followed by privacy and data-security risks (14) and sexual risks (8); education and awareness is the most recommended safety strategy (41 studies), with mediation and support systems second (21). The authors claim that culture was the most emphasized social factor (38 studies), and they report that only two studies engaged LGBTQIA+ youth or youth with disabilities and only one addressed politics and policy. They read this pattern as evidence that the field needs methodological diversification and a culturally situated, youth-centered research agenda.","pith_inferences":["Beyond the paper: because the corpus was built from three large bibliographic databases with an English/Spanish filter, the culture-emphasis finding is best read as a statement about indexed literature; a multilingual, regionally grounded search is the direct next test.","Beyond the paper: the review's counts imply a testable framing hypothesis—safety messages built on family honor or community responsibility may outperform individual-resilience messages in collectivist contexts—which the review does not itself test.","Beyond the paper: the two-study figure for marginalized youth may understate their presence if studies collected such demographics without foregrounding them; a full-text demographic-flagging pass would sharpen the claim."],"forward_implications":["If the review is correct, the evidence base for online safety in the Global South is strongest for survey-based prevalence of cyberbullying in Asia, so interventions built on this literature should be treated as region-specific rather than globally generalizable.","Researchers seeking to design culturally responsive safety tools should expect that education and awareness programs and family-based mediation are the two strategies with the most evidentiary support, while policy and regulation are a largely open question.","The near absence of LGBTQIA+, disabled, and other marginalized youth from the corpus means current safety guidance cannot yet claim to be inclusive, and future studies will need recruitment and measurement practices that record these identities.","The methodological imbalance predicts that qualitative and youth-centered studies will add the most new knowledge, not by replacing surveys but by explaining why cultural norms produce the patterns the surveys detect."],"supporting_citations":[{"why":"the earlier systematic review that documented the Global North focus this study extends","marker":"[98]"},{"why":"the systematic-review reporting guideline that structures the protocol","marker":"[82]"},{"why":"the updated reporting guideline behind the screening and selection flow diagram","marker":"[95]"},{"why":"the qualitative analysis method used to code the corpus into themes","marker":"[29]"},{"why":"the taxonomy of HCI contribution types used as an inclusion criterion","marker":"[135]"},{"why":"the country classification that defines the Global South for the search query","marker":"[41]"},{"why":"the qualitative-coding norms used to justify consensus coding without inter-rater reliability","marker":"[78]"}],"fun_headline_variants":["Global South youth safety research skews Asia and surveys","Cyberbullying focus leaves Global South youth safety gaps","Marginalized youth missing in Global South online safety research","Survey-heavy, Asia-centric: Global South youth safety studies","Online safety for Global South youth needs broader lens"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the 66 papers returned by the three chosen databases, the 2014–2024 window, and the English/Spanish language filter fairly represent the entire field of youth online safety research in the Global South; if regional, non-indexed, or non-English literatures are substantial, the observed Asia-heavy and survey-heavy pattern and the apparent silence on marginalized youth and policy could be sampling artifacts rather than properties of the field.","fun_headline_variants_meta":{"raw":{"variants":["Global South youth safety research skews Asia and surveys","Cyberbullying focus leaves Global South youth safety gaps","Marginalized youth missing in Global South online safety research","Survey-heavy, Asia-centric: Global South youth safety studies","Online safety for Global South youth needs broader lens"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00036,"raw_usage":{"total_tokens":1935,"prompt_tokens":924,"completion_tokens":1011,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":540,"completion_tokens_details":{"reasoning_tokens":934}},"tokens_in":540,"tokens_out":1011,"duration_ms":10132,"temperature":1.0,"reasoning_tokens":934,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:32:03.602285+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the review's Boolean query in regional and non-English publication venues that were not among the three selected databases, and code the additional hits with the same where/who/how/what scheme; if the added set contains many qualitative, policy-focused, or Latin American and African studies, the claims about regional concentration, methodological dominance, and the absence of marginalized-youth research would fail as statements about the field. A second check is to re-code all 66 full texts for whether any demographic or risk data about LGBTQIA+ or disabled youth appear even when not declared in the title or abstract.","supporting_citations":[],"review_version":1}