{"id":"1f9f9d9a-814a-429c-b3ba-4208e85ea908","arxiv_id":"2502.02459","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A systematic review of 206 smart ring papers that organizes the field into a taxonomy of interaction input, interaction output, passive in-body sensing, and passive out-body sensing.","lead":"This paper reviews 206 research papers on smart rings and sorts them into a four-part taxonomy covering input, output, health sensing, and activity sensing. It maps what rings can do today and points researchers to gaps such as open development platforms and raw sensor access.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Category counts in the taxonomy do not sum to 206, and several subsection counts contradict their parent totals, undermining the reported distribution that anchors the survey's central claim.","rationale":"The reader identified search coverage as the weakest assumption, but the paper's own reported numbers reveal a more direct and concrete problem. The abstract's central claim is that 206 publications were collected and categorized into four main areas. The four top-level category counts in Figure 2 sum to 215, not 206, and several subcategory counts similarly exceed their parent totals. These inconsistencies are not about external literature coverage; they are internal arithmetic contradictions in the paper's primary quantitative contribution. If the counts cannot be reconciled, then RQ1's distribution analysis and the taxonomy's per-category statistics are unreliable, which weakens the claim of a comprehensive and systematic map. The taxonomy itself may still be valuable, and the issue is likely fixable with a careful recount or explicit overlap reporting, so the appropriate verdict is CONDITIONAL rather than rejection. The concern is concrete and testable: recomputing counts from the paper's own table would resolve whether the discrepancy is a presentation error or a substantive classification flaw. The reader's assumption about search databases is also legitimate, but it is secondary to the paper's internal inconsistency, which should be addressed before the survey's quantitative claims can be accepted.","tokens_in":45087,"tokens_out":2868,"duration_ms":28775,"concrete_test":"Obtain the complete list of 206 included papers (the 63-column Google Sheets table referenced in Section 2) and recompute the count for each leaf and internal node in Figure 2 using the paper-level labels. If the recomputed totals equal 206 with single-label assignment, identify the specific papers causing the current discrepancies and correct the figure. If multi-label assignment is intended, report the number of papers in each intersection and state the overlap explicitly. A simpler first check: add the four top-level N values in Figure 2; if they exceed 206, the taxonomy is either overlapping or miscounted.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The most load-bearing concern is internal inconsistency in the paper's own category counts, not search coverage. Figure 2 lists N=116+29+46+24=215 papers, yet the paper claims 206 publications. Section-level counts also disagree: gesture subsections sum to 74 vs the stated N=66; trajectory subsections sum to 27 vs N=25; passive in-body subsections sum to 50 vs N=46; passive out-body subsections sum to 27 vs N=24. The methodology in Section 2 does not state whether papers may be assigned to multiple categories. If multi-label assignment is allowed, the sum can exceed 206 only if overlaps are explicitly identified, but no overlap analysis is provided. If single-label assignment is used, the counts are arithmetically wrong. Either way, the distributional claims in Figure 2 and Section 8, which are central to RQ1 and to the taxonomy's empirical grounding, are not reproducible from the reported numbers.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a systematic literature review of smart ring research, reporting a final corpus of 206 publications collected from ACM Digital Library and IEEE Xplore plus backward chaining. It proposes a four-layer taxonomy (Application, Phenomena, Fundamental Phenomena, Sensors) and organizes the literature into four application areas: interaction-input, interaction-output, passive sensing of in-body features, and passive sensing of out-body features. For each area it reviews sensing modalities, actuators, applications, and evaluation practices, and it closes with challenges and future research directions.","tokens_in":45224,"tokens_out":2598,"duration_ms":27274,"significance":"If the reported corpus and counts are made internally consistent, this would be a timely and useful contribution. It updates older surveys (Rissanen et al. 2013; Shilkrot et al. 2015) and broadens the scope of Vatavu and Bilius (2021) by covering not only gesture input but also haptic output, physiological sensing, activity recognition, and authentication. The phenomena-to-fundamental-phenomena flow in Section 3.2 is a valuable organizing device, and the paper explicitly credits prior taxonomies from Shilkrot et al. and Röddiger et al. The main empirical anchor of the survey, however, is the distribution of papers across categories, and the reported counts are not internally consistent; this currently prevents the reader from trusting the principal descriptive claims.","major_comments":[{"comment":"The category counts in Figure 2 do not sum to the claimed corpus size. The four top-level categories give 116 + 29 + 46 + 24 = 215, while the abstract and Section 2 state that the final set contains 206 papers. Section 2 does not state whether a paper may be assigned to multiple categories. If multi-label assignment is allowed, the overlap must be quantified and described; if single-label assignment is used, the arithmetic is wrong. This is load-bearing because the claimed distribution is the main evidence for RQ1 and for the taxonomy's empirical grounding.","section":"§3.1, Figure 2 and Abstract"},{"comment":"Several subsection counts contradict their parent counts. In Figure 5, Gesture is labeled N=66 but the listed subcategories sum to 26+6+9+5+3+25=74, and Trajectory is labeled N=25 but its subcategories sum to 3+13+8+3=27. In Figure 9, Activity Recognition is labeled N=8 but Full-body Activity (N=2) and Hand Activity (N=8) sum to 10. These inconsistencies are too large to be rounding or typographical noise, and they directly affect the taxonomy's structure as presented in the main outline.","section":"§4.1, Figure 5 and §7.1, Figure 9"},{"comment":"The passive sensing - in-body feature section has a similar accounting problem. Figure 8 labels the top-level category as N=46, but the subcategories sum to 18 (Physiological Sensing) + 17 (Health Tracking) + 13 (Diagnoses and Treatment) + 2 (Biometric Authentication) = 50. Section 8.2 also relies on such counts (e.g., 'haptic feedback (26 out of 29)'), so the internal inconsistency propagates into the discussion. The authors should re-audit every count in Figures 2, 5, 8, and 9 and state the assignment rule (single-label vs. multi-label) explicitly in Section 2.","section":"§6, Figure 8"}],"minor_comments":[{"comment":"There are several language issues in the methodology section, e.g., 'we perfomed' and 'In total, we this results in 593 relevant publications'. These should be corrected in revision.","section":"§2"},{"comment":"The search is limited to ACM Digital Library and IEEE Xplore, supplemented by backward chaining. This is a defensible choice, but the authors should explicitly acknowledge the risk of missing venues and non-English publications, and explain why the backward-chaining step is expected to mitigate that risk.","section":"§2"},{"comment":"'IEEE Explore' should be 'IEEE Xplore'.","section":"§2"},{"comment":"The caption states that one paper in the interaction-input category was published in 1997, but the figure axes are labeled 2000 to 2024. The authors should clarify how the 1997 paper is displayed.","section":"Figure 1"},{"comment":"The price column mixes different currencies (e.g., $, €) without a consistent notation. A footnote explaining currency and approximate conversion would improve readability.","section":"Table 6"}],"recommendation":"major_revision","confidential_remarks":"The paper is not beyond repair, but the internal count inconsistencies are serious because they affect the paper's central descriptive claim. I would ask the editor to require the authors to provide a supplementary artifact (e.g., a table listing every included paper and its category assignment) so that all counts can be verified. If the counts are corrected and the multi-label rule is stated, I would view the manuscript favorably."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a useful and mostly well-built survey of smart ring research, and it deserves a serious referee even though I would not accept it as is. The central count data have arithmetic problems that need fixing before any distributional claims can be trusted.\n\nWhat's genuinely good: it covers 206 papers (2000-2024), which is broader than prior surveys, and organizes the field into four application areas with a phenomena-to-sensor flow that actually helps a reader see design space. The inclusion/exclusion criteria, two-stage screening, and backward chaining are described well enough to be audited. The tables—sensors and fundamental phenomena, commercial health rings, ground truth methods, gesture frequency—are useful reference material.\n\nWhere it's soft: the numbers don't add up. Figure 2's four top categories sum to 215, not 206. Within interaction-input, Gesture is listed as N=66 but its six subsections sum to 74; Trajectory is N=25 but its four subsections sum to 27. In passive in-body, the four main subsections sum to 50 against stated N=46. In passive out-body, they sum to 27 against N=24. Some of these could be multi-label classification, but the methodology never says so, no overlap analysis is offered, and the taxonomy is described as a binary decision tree. As written, the distributional claims that answer RQ1 are not reproducible from the paper's own numbers. That's a load-bearing flaw for a systematic review, even if the qualitative taxonomy and discussion remain solid.\n\nThe two-database search is a lesser concern; backward chaining plus the venues covered make it defensible. I'd also want the authors to clarify the 87% sensing statistic, which appears without derivation.\n\nWho it's for: smart ring and wearable HCI researchers, especially people looking for a structured map and gap list. A careful reader will get real value from the taxonomy and tables despite the count issues.\n\nRecommendation: send it to peer review, but require the authors to reconcile the category counts and either state a clear multi-label policy with overlap data or correct the arithmetic. This is fixable without new experiments.","headline":"A genuinely useful survey whose central distributional claims are undermined by arithmetic inconsistencies that should be fixed before publication.","tokens_in":45811,"tokens_out":2244,"would_cite":true,"duration_ms":22334,"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 systematic review of 206 smart-ring papers organizes the field into four application areas connected by a phenomena-to-sensor taxonomy.","keywords":["smart rings","wearable computing","finger-worn","finger-mounted","finger-attached","finger augmentation","systematic literature review","gesture recognition"],"falsifier":"Run the same inclusion protocol with additional scholarly databases and broader keyword variants; if that search surfaces a substantial number of relevant English-language ring papers from 2000 to 2024 that the current 206-paper set omits, the claimed comprehensiveness fails. A second check is to re-run the two-author screening on the 593 initially retrieved records and see whether a comparable review converges on the same 159-plus-47 paper selection.","tokens_in":44889,"feed_emoji":"💍","tokens_out":6721,"duration_ms":68157,"temperature":0.7,"pith_summary":"Smart rings are small enough to wear all day and sit where the hand is most expressive, so they can serve as gesture controllers and as quiet health monitors at the same time. This paper tries to establish what that field actually contains by systematically reviewing 206 smart-ring publications and sorting them into four application families: interaction as input, interaction as output, passive sensing of in-body features, and passive sensing of out-body activity. It further argues that every application can be traced down a chain from application to sensed phenomenon to fundamental phenomenon to sensor, so that disparate ring systems are compared by the physical signals they exploit. A sympathetic reader would care because the review gives researchers a shared map of what rings can do, which sensors are behind each capability, and where the open design space is.","feed_headline":"206 smart-ring papers, sorted into one taxonomy","feed_subtitle":"A systematic review links each ring application to the sensors and physical signals behind it.","key_machinery":"The load-bearing mechanism is the four-layer taxonomy of application, phenomena, fundamental phenomena, and sensors, modeled on the structure used for earable computing. Fundamental phenomena are defined as what a sensor can directly observe, such as blood perfusion, body resistance, motion, temperature, or emitted acoustic and laser signals; all higher-level phenomena in the review, from gesture to heart rate to drink composition, are derived from these primitives. A companion binary decision tree classifies applications by user intention, splitting first into interaction versus passive sensing and then into input/output and in-body/out-body categories. Together the two structures let the review compare 206 papers on a common grid and expose which sensing primitives already have mature applications and which remain underexplored.","core_discovery":"The central claim is that smart ring research can be organized into a single taxonomy with two levels: a binary decision tree over user intention (active interaction versus passive sensing, then input/output and in-body/out-body) and a four-layer sensing chain of application, phenomena, fundamental phenomena, and sensors. The review places the 206 collected papers in this structure, showing that most work concentrates on gesture and trajectory input, that haptic vibration dominates output research, and that physiological monitoring relies heavily on a small set of commercial rings. It also enumerates the fundamental phenomena, such as motion, blood perfusion, body resistance, emitted sound, and visual appearance, that sensors directly capture and from which higher-level applications are inferred. If the taxonomy is right, the field's scattered prototypes and products become commensurable: any ring system can be located by what it senses, what it derives, and what it lets users do.","pith_inferences":["If the taxonomy generalizes, it could be used prospectively: classify an unreleased ring's sensors and intended use, and the map predicts which application families it can enter and which phenomena it can support.","The field's literature total is probably a lower bound, since commercial activity tracking and proprietary health algorithms are often documented outside peer-reviewed venues; the four-category counts may therefore underrepresent passive sensing.","The clean separation between interaction and passive sensing is a review-time convenience; real products will increasingly combine both, and a merged category might be where the next generation of rings is built.","A testable extension would be to map the taxonomy onto smartwatches or earables; the same application-to-fundamental-phenomena chain could reveal which finger-worn capabilities are genuinely unique to rings."],"forward_implications":["Gesture and trajectory input form the largest cluster of ring research, so future input work can build on a cataloged set of gesture types rather than starting from scratch.","Because fundamental phenomena are few and reusable, a designer can choose a sensor by asking which directly observable signal a new application needs, rather than by copying an existing device.","Health-oriented ring studies depend heavily on commercial products such as the Oura Ring, and the review's finding that raw sensor data is often inaccessible implies that validation and custom algorithms will require more open platforms.","Haptic output is almost the entire output category, and the absence of systematic ring-versus-watch or ring-versus-glove comparisons marks a concrete gap for future studies."],"supporting_citations":[{"why":"Supplies the early survey of ring-shaped interfaces that this review updates.","marker":"[146]"},{"why":"Defines finger augmentation devices and the ring form-factor used to bound the corpus.","marker":"[157]"},{"why":"Provides the prior systematic review of ring gesture input and the form-factor definitions this review extends.","marker":"[177]"},{"why":"Contributes the four-layer taxonomy and the systematic literature review method adapted here.","marker":"[147]"},{"why":"Offers the working definition of ring gestures that the paper adapts into a broader smart-ring definition.","marker":"[38]"},{"why":"Guides the systematic literature review process for research questions and search steps.","marker":"[1]"},{"why":"Guides the iterative SLR approach that shapes the methodology.","marker":"[86]"},{"why":"Supplies the fine-to-coarse gesture taxonomy reused in the interaction-input section.","marker":"[96]"},{"why":"Provides the 21-DoF hand pose model and an open ring platform cited in the taxonomy and hand-pose tracking.","marker":"[218]"}],"fun_headline_variants":["Smart-ring literature mapped into a single taxonomy","206 smart ring papers organized into one taxonomy","A systematic map of smart ring research","Smart rings decoded: a taxonomy for 206 papers","The smart ring field, ordered by sensing and intent"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The review's coverage depends on the assumption that searching two major digital libraries with the chosen keywords, filtered by an English-only, peer-reviewed, evaluation-required criterion, and then backward-chaining from reference lists, captures essentially all relevant smart-ring research.","fun_headline_variants_meta":{"raw":{"variants":["Smart-ring literature mapped into a single taxonomy","206 smart ring papers organized into one taxonomy","A systematic map of smart ring research","Smart rings decoded: a taxonomy for 206 papers","The smart ring field, ordered by sensing and intent"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000606,"raw_usage":{"total_tokens":2797,"prompt_tokens":887,"completion_tokens":1910,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":503,"completion_tokens_details":{"reasoning_tokens":1841}},"tokens_in":503,"tokens_out":1910,"duration_ms":15055,"temperature":1.0,"reasoning_tokens":1841,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T12:02:53.768449+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same inclusion protocol with additional scholarly databases and broader keyword variants; if that search surfaces a substantial number of relevant English-language ring papers from 2000 to 2024 that the current 206-paper set omits, the claimed comprehensiveness fails. A second check is to re-run the two-author screening on the 593 initially retrieved records and see whether a comparable review converges on the same 159-plus-47 paper selection.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines finger augmentation devices and the ring form-factor used to bound the corpus."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the prior systematic review of ring gesture input and the form-factor definitions this review extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Contributes the four-layer taxonomy and the systematic literature review method adapted here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the 21-DoF hand pose model and an open ring platform cited in the taxonomy and hand-pose tracking."}],"review_version":1}