{"id":"109700db-a3f8-4e0d-94ef-86f3d3594ff8","arxiv_id":"2504.21477","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"This survey synthesizes 110 studies on electrotactile and electrical muscle stimulation haptics in HCI, showing a strong focus on the forearm and identifying key challenges in standardization, calibration, and adaptive rendering.","lead":"Systematic review of 110 papers on electrical stimulation haptic feedback in human-computer interaction, covering devices, perception, comparisons with other feedback modalities, and applications. It finds the field concentrated on the upper limbs and identifies standardization, skin impedance variability, and closed-loop rendering as the main open gaps.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corpus completeness and the unpublished 110-study list are the load-bearing weakness for the survey's quantitative claims.","rationale":"The reader's weakest-assumption identification is correct: the survey's central quantitative claims are only as strong as the completeness and codability of the 110-paper corpus. My stress-test pass did not find a more severe internal inconsistency. The PRISMA procedure is described reasonably, the synthesis is consistent with the cited literature, and the body-site percentages are internally plausible given the table, though the multi-site classifications in Table III would ideally be clarified. The main vulnerability is that the corpus list and extraction data are not published, and the database selection omits major non-ACM/IEEE HCI publication channels. Because the survey explicitly aims to be 'comprehensive' and derives quantitative distributions and gap analyses from the corpus, this is load-bearing: selection bias could change the 47.3% forearm figure and the conclusion that full-body feedback is underexplored. The proposed concrete test, an independent Scopus/Web of Science validation search plus release of the extraction data, would settle the concern. This does not warrant moving from the reader's CONDITIONAL verdict to ACCEPT or REJECT, because the concern is about verifiability and potential bias, not about demonstrated factual error in the synthesis.","tokens_in":36705,"tokens_out":8987,"duration_ms":93828,"concrete_test":"Perform an audit by (1) releasing the full list of 110 included DOIs together with the DE5 stimulation-site classifications and the PRISMA numbers per database, and (2) independently running the finalized keyword string ('electrotactile' OR 'electro tactile' OR 'electrocutaneous' OR 'electro-cutaneous' OR 'electro cutaneous' OR 'electrotactile stimulation' OR 'electrical tactile feedback' OR 'electric haptic feedback' OR 'electrical feedback' OR 'electric stimulation' OR 'electrical muscle stimulation' OR 'electric tactile feedback') in Scopus or Web of Science for 2014-2024, restricted to HCI-relevant venues, and classifying all additional relevant studies by body site using the same inclusion criteria. Recompute the forearm percentage and the number of non-upper-limb studies.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The survey's central contribution is a set of quantitative corpus-level claims: the forearm is the most studied site (47.3%, n=52), the fingertip follows (20.0%), and full-body electrotactile feedback is underexplored. These percentages are computed over a corpus assembled from only ACM Digital Library, IEEE Xplore, and six selected journals (Section II-B), using eligibility criteria that require the study to 'explore how electrical haptic feedback enhances user experience' (Section II-C). The paper reports PRISMA flow counts (543 records, 110 included) but does not publish the list of the 110 included studies or the DE1-DE15 extraction data, including the DE5 stimulation-site codings that directly produce the 47.3% figure. Without this list, a reader cannot verify that the corpus is complete, that the site classifications are consistent, or that the gap analysis (e.g., 'full-body feedback remains largely unexplored') is not an artifact of the database selection. In particular, relevant HCI work on electrical haptic feedback appears in journals published by Springer, Elsevier, Wiley, and SAGE (e.g., Virtual Reality, International Journal of Human-Computer Studies, Human Factors), which are not in the search strategy. If those venues contain a different body-site distribution, the headline percentages and the underexplored-region conclusion would shift. The paper itself does not flag this coverage limitation or provide a reproducibility appendix, so the 'comprehensive' claim is currently unsupported in a way that matters for all downstream synthesis.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a systematic review of electrical stimulation haptic feedback in human-computer interaction, following the PRISMA 2020 reporting guideline. The authors searched ACM Digital Library, IEEE Xplore, and six selected journals, screened 543 records down to 110 included studies, and extracted 15 data items (DE1-DE15) covering devices, stimulation parameters, perception, multisensory comparisons, and applications. The paper's central quantitative claims are that the forearm is the most studied stimulation site (47.3%, n=52), that typical tactile stimulation parameters are 0-6 mA intensity, 0-400 µs pulse width (written as \"0-400 s\" in the text), and 0-300 Hz frequency, and that hybrid multimodal feedback generally outperforms unimodal feedback in accuracy and immersion. It also identifies full-body electrotactile feedback as an underexplored area and proposes future directions including multimodal integration, high-resolution devices, adaptive stimulation, and full-body systems.","tokens_in":36903,"tokens_out":4892,"duration_ms":53181,"significance":"If the quantitative corpus-level statements are reproducible, this survey would provide a useful baseline for researchers and developers in electrotactile HCI. The paper's strengths include a clearly documented PRISMA flow, a structured extraction rubric (DE1-DE15), quantitative mapping of body sites and stimulation parameters, and a systematic comparison of feedback modalities. The discussion of demographic limitations and open-loop rendering is also a useful contribution. However, the corpus list and extraction data are not released, and the search scope excludes several venues that publish HCI haptics work; as a result, the headline percentages and the gap analysis are not independently auditable. The central value of the paper therefore depends on transparency measures that are currently missing.","major_comments":[{"comment":"The search strategy is limited to ACM Digital Library, IEEE Xplore, and six selected journals (Nature Communications, Nature Electronics, npj Flexible Electronics, Nature Machine Intelligence, Science Advances, Science Robotics). Many HCI venues that publish electrical haptic feedback work, such as Virtual Reality (Springer), International Journal of Human-Computer Studies (Elsevier), and Human Factors (SAGE), are not included. Since the paper's headline claims are corpus-level distributions (e.g., the 47.3% forearm share and the statement that full-body feedback is underexplored), this coverage restriction can change those conclusions; the paper neither justifies the completeness of this set nor reports a sensitivity analysis. I recommend either expanding the search or explicitly documenting this coverage limitation as a bound on the claims.","section":"Section II-B, Section II-C"},{"comment":"The manuscript does not publish the list of the 110 included studies or the DE1-DE15 extraction data. As a result, the percentages in Section III-A, including the 47.3% forearm figure, cannot be independently audited: the reader cannot check which studies were coded to which DE5 location or verify the screening decisions. A supplementary file with the full corpus and the extraction table should be provided; at minimum, the per-DE5 counts per study and the exclusion decisions should be released so that the corpus-level claims can be reproduced.","section":"Section II-D, Section II-E"},{"comment":"The paper's key quantitative parameter summary is not fully reproducible. The pulse-width range is reported as \"0-400 s\" both in Section III-C and in the Discussion, but the unit should be microseconds (µs), as correctly stated earlier in Section I.C.2. In addition, the density plots do not report the number of studies contributing to each plot; because the text says only studies with explicitly reported data were included, the reader cannot judge how representative the reported ranges are. Please fix the unit and add the sample size (N) for each plot, ideally with the underlying extracted values.","section":"Section III-C, Figure 6"},{"comment":"Eligibility criterion (3) excludes studies that do not address user experience, such as purely technical implementations or device development without experimental or user data. This is a defensible scope choice, but it directly shapes the device and body-site gap analysis: hardware-only electrotactile papers on underexplored body regions would be excluded by construction, which reinforces the conclusion that full-body feedback is underexplored. The authors should acknowledge this bias and, if possible, quantify how many technical-only papers were excluded and at which body sites, since this bears directly on the comprehensiveness claim.","section":"Section II-C, Section III-A, Section IV.C.4"}],"minor_comments":[{"comment":"The keyword list contains the term \"electrotactile\" twice; please remove the duplicate.","section":"Section II-B"},{"comment":"Reference [99] lists the author as \"M. DrAlonzo\"; this should likely be \"M. D'Alonzo\" and should be checked against the original publication.","section":"References"},{"comment":"The caption says \"Publications Per Y ear\" (typo for \"Year\"), and the figure plots only venues with annual publication counts of at least two; the caption should state this inclusion rule and, ideally, provide the remaining venues in a table for completeness.","section":"Figure 3"},{"comment":"The text at Section II.E describes DE9-DE12 as documenting \"waveform, frequency, bandwidth, and intensity,\" but Table II labels DE11 as \"Stimulation Duration\" and does not use \"bandwidth\"; please align the terminology to avoid confusion.","section":"Section II-E, Table II"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of a CS/HCI journal and the systematic methodology is essentially sound. The main obstacles to acceptance are transparency and reproducibility: the corpus and extraction data are not released, the search coverage is narrower than the \"comprehensive\" claim implies, and a headline quantitative result contains a unit error. I found no evidence of misconduct; the self-citation pattern is modest and relevant. I would support acceptance after these issues are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuinely useful survey of 110 electrotactile/EMS papers, with a sensible analytic split between tactile and kinesthetic effects and some honest gap analysis. It is not a breakthrough, but it is a solid map of the subfield that deserves to be published after moderate revision.\n\nWhat's new relative to Zhou et al. 2022 and Kourtesis et al. 2022: the 2014–2024 window, the body-site distribution numbers (forearm 47.3%, fingertip 20%), the density plots of stimulus parameters, and the explicit separation of EMS-induced kinesthetic effects. The PRISMA flow is described carefully, and the extraction rubric (DE1–DE15) is a useful methodological contribution by itself. I also appreciate that the survey flags standardization and calibration as open problems without pretending there is a consensus fix.\n\nThe soft spots are real but not fatal. The biggest one is that the authors do not publish the list of the 110 included studies or the DE1–DE15 extraction data. That makes the headline percentages and the 'full-body remains underexplored' conclusion impossible to audit. Combined with a search restricted to ACM, IEEE, and six high-profile journals, the 'comprehensive' label is too strong. There is certainly relevant work in Springer, Elsevier, Wiley, and SAGE venues (Virtual Reality, IJHCS, Human Factors). The authors should either release the full corpus and extraction tables as supplementary material or soften the coverage claim. This is a fixable problem — not a reason to reject.\n\nThe unit typos in the density-plot axis labels ('0-400 s' for pulse width) are minor but embarrassing and should be caught in copyedit. The claim that hybrid multimodal feedback outperforms unimodal is a reasonable synthesis of the cited comparisons, not an overreach, though it rests on a small number of studies with different tasks.\n\nThe math in Section I.B is background only, taken from Kajimoto and Rattay; it is not a load-bearing part of the survey. Self-citations are present but not egregious.\n\nVerdict: worth a serious peer review. The methodological choices are defensible; the missing corpus data is the one thing that must be fixed before publication. If the authors add a reproducibility appendix, this becomes a reference worth citing. I'd bring it to a reading group only to discuss the transparency of systematic reviews in HCI, not for the substance.","headline":"Useful, well-structured survey of electrotactile HCI whose corpus data needs to be released before the 'comprehensive' claim is credible.","tokens_in":37523,"tokens_out":1595,"would_cite":true,"duration_ms":16574,"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 110 studies maps electrical haptic feedback in human-computer interaction, finding the field concentrated on the forearm and reporting that hybrid multimodal feedback outperforms unimodal feedback.","keywords":["electrical haptic feedback","electrotactile stimulation","electrical muscle stimulation","haptic perception","multimodal feedback","wearable haptic devices","virtual reality","systematic review"],"falsifier":"Re-run the search in a broader set of bibliographic databases and reconstruct the corpus from the stated keyword set and eligibility criteria. If the stimulation-site distribution changes materially, for example the forearm share falls well below 47.3% when additional leg or torso studies are included, the review's gap analysis would need revision.","tokens_in":1610,"feed_emoji":"⚡","tokens_out":1816,"duration_ms":74697,"temperature":0.7,"pith_summary":"This paper aims to establish a reliable, systematic map of electrical haptic feedback in human-computer interaction by analyzing 110 research papers from 2014 to 2024. It claims that the literature supports specific quantitative baselines: the forearm is the most studied site (47.3% of papers), typical tactile stimulation uses intensities of 0–6 mA, pulse widths of 0–400 microseconds, and frequencies of 0–300 Hz, and hybrid multimodal feedback generally beats unimodal feedback in accuracy and immersion. If the map is correct, designers and researchers gain a common starting point for device design, parameter calibration, and gap-filling research, with full-body feedback identified as the clearest open area.","feed_headline":"Forearm dominates electrical touch research, survey finds","feed_subtitle":"110-study review pinpoints typical stimulation ranges and finds hybrid multimodal feedback beats unimodal.","key_machinery":"The argument is carried by the corpus itself plus a structured 15-field data extraction rubric (DE1–DE15). Each of the 110 included studies is coded for objective, stimulation location, device type, electrode type, waveform, frequency, duration, amplitude, participants, experimental method, and conclusions; those fields feed the distribution tables, density plots, and qualitative comparisons that produce the paper's quantitative claims. The stratification by stimulation site and the density plots of stimulation intensity, pulse width, and frequency are the mechanism that turns a list of papers into the 47.3% forearm share and the 0–6 mA / 0–400 microsecond / 0–300 Hz baselines.","core_discovery":"The central claim of the paper is that a systematic reading of 110 studies from 2014 to 2024 supports a quantitative description of electrical haptic feedback in human-computer interaction. The forearm is the dominant stimulation site (47.3% of the corpus), followed by the fingertip (20.0%), upper arm (14.5%), finger (9.1%), and wrist (8.2%); effective tactile stimulation clusters at 0–6 mA, pulse widths of 0–400 microseconds, and frequencies of 0–300 Hz; and in direct comparisons, hybrid multimodal feedback combining electrotactile stimulation with vibrotactile, force, or thermal cues generally achieves higher accuracy and immersion than unimodal feedback. The paper further claims that electrical stimulation can elicit both tactile sensations such as texture, pressure, shape, and temperature, and kinesthetic or pseudo-kinesthetic effects through muscle activation, and it identifies the main bottlenecks as device standardization, reliance on subjective reports, open-loop rendering, skin-impedance variability, manual calibration, and limited participant demographics.","pith_inferences":["Because the review is restricted to a specific set of literature databases, one direct test of its distribution claims would be to search additional bibliographic databases and see whether the forearm's 47.3% share holds.","The reported lack of AI-driven tactile rendering suggests a concrete next step: closed-loop systems that adapt pulse parameters in real time to measured skin impedance or muscle response.","The upper-limb concentration may partly reflect convenience of experimentation rather than where the technology performs best, so body sites with different receptor density and skin impedance could behave quite differently from the reported averages.","If hybrid multimodal feedback is as consistently superior as the review reports, a sensible benchmark for future electrotactile studies would be to compare new systems against the best available unimodal baseline rather than against no feedback at all."],"forward_implications":["A designer starting an electrotactile project can use 0–6 mA, 0–400 microsecond pulses, and 0–300 Hz as first-guess ranges rather than starting from scratch.","Researchers should expect that claims about body sites outside the upper limb rest on far fewer studies than the forearm and fingertip, so those areas need dedicated empirical work.","Hybrid systems that pair electrotactile stimulation with vibrotactile, force, or thermal cues are the more promising path for accuracy and immersion, according to the reviewed comparisons.","The field needs shared benchmarks and standardized reporting of electrode type, waveform, and stimulation parameters before cross-study comparisons become reliable.","Full-body electrotactile systems are an open opportunity, since the reviewed literature offers little coverage beyond the hands, arms, and wrists."],"supporting_citations":[{"why":"Supplies the systematic review reporting protocol that structures the screening and inclusion process.","marker":"[53]"},{"why":"Frames electrotactile perception properties and applications, serving as the prior review that motivates this broader survey.","marker":"[5]"},{"why":"Previous systematic review focused on hand and arm electrotactile applications, providing the baseline this paper extends to the whole body.","marker":"[52]"},{"why":"Provides the electrical stimulation model and design principles used to explain how electrodes activate nerves.","marker":"[23]"},{"why":"Underlying bioelectric model of nerve membrane activation that the stimulation model builds on.","marker":"[24]"},{"why":"Reference for the metaverse vision of full-body haptic engagement, used to motivate the gap analysis.","marker":"[164]"},{"why":"Introduces event-related potentials as an objective measure of visuo-haptic mismatch, supporting the paper's claims about multimodal immersion and objective evaluation.","marker":"[95]"},{"why":"Hybrid vibro-electrotactile interface achieving high recognition accuracy, a key piece of evidence for the hybrid multimodal advantage.","marker":"[99]"},{"why":"Establishes a parameter-intensity model for perceived intensity, supporting the review's quantitative treatment of stimulation parameters.","marker":"[155]"},{"why":"Documents skin impedance nonlinearity and PWM/Kalman stabilization, supporting the paper's claims about calibration and impedance variability.","marker":"[144]"}],"fun_headline_variants":["Electric touch: forearm leads, hybrid wins","110 studies: forearm dominates, hybrid beats unimodal","Forearm is the hotspot for electric haptics; hybrid is best","Survey of 110 papers: forearm is top electric touch target"],"cache_read_input_tokens":39552,"weakest_assumption_plain":"The whole picture rests on the assumption that the chosen literature databases plus a short list of six journals capture essentially all relevant 2014–2024 work on electrical haptic feedback; if substantial relevant studies sit in other bibliographic databases, the reported percentages, including the 47.3% forearm share, could be skewed.","fun_headline_variants_meta":{"raw":{"variants":["Electric touch: forearm leads, hybrid wins","110 studies: forearm dominates, hybrid beats unimodal","Forearm is the hotspot for electric haptics; hybrid is best","Survey of 110 papers: forearm is top electric touch target"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001196,"raw_usage":{"total_tokens":4901,"prompt_tokens":880,"completion_tokens":4021,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":496,"completion_tokens_details":{"reasoning_tokens":3954}},"tokens_in":496,"tokens_out":4021,"duration_ms":27689,"temperature":1.0,"reasoning_tokens":3954,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:01:48.249425+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the search in a broader set of bibliographic databases and reconstruct the corpus from the stated keyword set and eligibility criteria. If the stimulation-site distribution changes materially, for example the forearm share falls well below 47.3% when additional leg or torso studies are included, the review's gap analysis would need revision.","supporting_citations":[],"review_version":1}