{"id":"dbddde92-3184-4613-9ecb-0a59f484b4d0","arxiv_id":"2605.22904","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Formalizes suicide risk assessment (SRA) from metro videos and benchmarks an interpretable pipeline with 83.2% ROC-AUC using tracking, activity recognition, segmentation, and risk heatmaps.","lead":"This paper introduces the first interpretable AI framework for assessing suicide risk from metro station surveillance videos by combining person tracking, activity recognition, platform segmentation, and trajectory risk modeling. A smart generalist might read it because it shows how computer vision can be applied to real-world suicide prevention in public spaces.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Reported 83.2% ROC-AUC rests on undefined ground-truth labeling of suicide-risk events in real surveillance data","rationale":"The reader's weakest_assumption directly flags the missing validation details; the load-bearing gap is precisely the absence of any description of how risk labels are obtained, which prevents any assessment of whether the reported metric supports the claim.","tokens_in":1757,"tokens_out":288,"duration_ms":20161,"concrete_test":"Extract the dataset and evaluation sections; confirm whether positive examples are tied to documented incidents or post-hoc video review, the number of such events, and whether labeling was performed blind to model outputs. If labeling protocol is absent or circular, recompute AUC on a held-out subset with independently verified labels.","verdict_should_be":"UNVERDICTED","load_bearing_attack":"The central claim is that the integrated pipeline (tracking + activity recognition + segmentation + trajectory risk heatmaps) produces a usable risk score, validated at 83.2% ROC-AUC on real data. This requires a concrete definition of positive and negative labels that is independent of the model's own outputs. No such definition, dataset construction protocol, or labeling source (incident logs, expert annotation criteria, temporal windows, etc.) is supplied. Without it, the AUC cannot be interpreted as evidence that the accumulated-evidence formulation measures the intended construct rather than proxy statistics.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper formalizes the task of Suicide Risk Assessment (SRA) from metro station surveillance videos as distinct from isolated subtasks or direct intent inference. It proposes an interpretable framework that accumulates evidence via person tracking, activity recognition, semantic segmentation of the platform, and trajectory-driven risk heatmap modeling, and reports benchmarking a complete operational pipeline that achieves 83.2% ROC-AUC on real surveillance data.","tokens_in":1841,"tokens_out":395,"duration_ms":28788,"significance":"If the validation details were supplied and the metric shown to be reliable, the work would be significant as one of the first end-to-end interpretable pipelines for this socially important application in computer vision. It correctly emphasizes accumulated behavioral and contextual cues over direct intent modeling and opens directions for AI for social good. The absence of any dataset construction, labeling protocol, or baseline comparison currently prevents assessing whether the result advances the state of the art.","major_comments":[{"comment":"Abstract (and the evaluation section it summarizes): the central claim that the pipeline 'achieves 83.2% ROC-AUC on real surveillance data' is load-bearing for the contribution, yet the manuscript supplies no information on dataset size, how ground-truth positive and negative suicide-risk events are defined independently of the model (e.g., incident logs, expert annotation criteria, temporal windows), train/test splits, cross-validation, baselines, or error analysis. Without an independent labeling protocol the reported AUC cannot be interpreted as evidence that the accumulated-evidence formulation measures the intended construct.","section":"Abstract and Evaluation"}],"minor_comments":[{"comment":"The abstract states the framework is 'the first' without referencing prior related work on behavior analysis or risk modeling in surveillance; a brief related-work paragraph would improve context.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback and for acknowledging the potential significance of formalizing suicide risk assessment as a distinct task. We address the major comment on evaluation details below.","responses":[{"response":"We agree that the manuscript currently provides insufficient details on the dataset and evaluation protocol. In the revised manuscript we will add a dedicated subsection describing: the size and characteristics of the real surveillance dataset; the independent ground-truth labeling protocol (including use of incident logs, expert annotation criteria, and temporal windows for positive/negative events); the train/test splits and any cross-validation procedure; relevant baselines; and error analysis. These additions will allow readers to properly interpret the 83.2% ROC-AUC in the context of the accumulated-evidence formulation.","revision_made":"yes","referee_comment":"[Abstract and Evaluation] Abstract (and the evaluation section it summarizes): the central claim that the pipeline 'achieves 83.2% ROC-AUC on real surveillance data' is load-bearing for the contribution, yet the manuscript supplies no information on dataset size, how ground-truth positive and negative suicide-risk events are defined independently of the model (e.g., incident logs, expert annotation criteria, temporal windows), train/test splits, cross-validation, baselines, or error analysis. Without an independent labeling protocol the reported AUC cannot be interpreted as evidence that the accumulated-evidence formulation measures the intended construct."}],"tokens_in":1363,"tokens_out":307,"duration_ms":37303,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The central point is that this work formalizes suicide risk assessment as a distinct task and assembles an operational pipeline from person tracking, activity recognition, platform segmentation, and trajectory heatmaps. That combination is presented as the first complete attempt at the problem rather than isolated subtasks, and the use of accumulated evidence plus visual heatmaps is a sensible choice for interpretability.\n\nThe paper does a reasonable job laying out why direct intent inference is avoided and how spatial-temporal cues can be aggregated. The abstract is clear on the intended components and the goal of supporting prevention in public transit.\n\nThe main weakness is the evaluation. The reported 83.2% ROC-AUC on real surveillance data is given without any description of dataset size, how positive and negative examples were defined or sourced, temporal windows, expert criteria, or even basic baselines. Without an independent ground-truth protocol, the metric does not show whether the pipeline measures the intended risk construct or simply correlates with other observable patterns. That gap is load-bearing for any claim of practical utility.\n\nThis paper is aimed at researchers working on applied computer vision for safety or social-good applications. Readers who want to see how standard CV modules can be wired together for a new end task may extract some value from the pipeline description. Anyone expecting reproducible performance numbers or a validated system will find the current version thin.\n\nI would send it for peer review because the application area matters and the accumulated-evidence framing is worth testing, but the authors would need to supply the missing labeling and validation details before the results can be assessed.","headline":"The paper outlines a multi-component pipeline for suicide risk scoring from metro video but the 83.2% ROC-AUC rests on an undefined labeling process for the positive cases, so the number cannot be interpreted yet.","tokens_in":2329,"tokens_out":405,"would_cite":false,"duration_ms":27026,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"An interpretable framework assesses suicide risk in metro stations from surveillance video by accumulating evidence through person tracking, activity recognition, platform segmentation, and trajectory heatmaps instead of inferring intent.","keywords":["suicide risk assessment","video surveillance","metro stations","person tracking","activity recognition","semantic segmentation","risk heatmap modeling","interpretable framework"],"falsifier":"A controlled test on new metro surveillance footage in which the framework produces risk scores that do not correlate with actual recorded incidents or intervention outcomes at rates above random chance.","tokens_in":2641,"feed_emoji":"🚇","tokens_out":465,"duration_ms":29562,"temperature":0.7,"pith_summary":"The paper formalizes suicide risk assessment as a distinct task in metro station surveillance video. It presents an interpretable pipeline that jointly processes person tracking, activity recognition, semantic segmentation of the platform, and trajectory-driven risk heatmap modeling to build risk scores from accumulated behavioral cues over time. The approach is tested on real data, reaching 83.2 percent ROC-AUC. A sympathetic reader would care because the method aims to support timely intervention in suicide prevention by handling the spatial context and temporal dynamics that isolated subtasks miss.","feed_headline":"Pipeline reaches 83.2% ROC-AUC on metro suicide risk video","feed_subtitle":"Tracks passengers, recognizes activities, segments platforms and builds heatmaps to accumulate evidence for prevention.","key_machinery":"The interpretable pipeline that integrates person tracking, activity recognition, semantic segmentation of the platform, and trajectory-driven risk heatmap modeling to accumulate behavioral evidence over time and produce risk scores.","core_discovery":"The authors introduce the first interpretable framework for Suicide Risk Assessment in metro stations that assesses risk from accumulated evidence by incorporating person tracking, activity recognition, semantic segmentation of the platform, and trajectory-driven risk heatmap modeling, achieving 83.2 percent ROC-AUC on real surveillance data rather than focusing on isolated subtasks or attempting to infer intent directly.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["Interpretable framework achieves 83.2% ROC-AUC in metro suicide risk assessment","Passenger tracking and heatmaps enable 83.2% ROC-AUC suicide risk video analysis","Activity recognition and platform segmentation reach 83.2% ROC-AUC for metro SRA","Interpretable pipeline benchmarks metro suicide risk at 83.2% ROC-AUC"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The integrated pipeline of tracking, recognition, segmentation, and heatmap modeling can reliably assess suicide risk from video without direct intent inference, and the reported ROC-AUC indicates practical utility for prevention.","fun_headline_variants_meta":{"raw":{"variants":["Interpretable framework achieves 83.2% ROC-AUC in metro suicide risk assessment","Passenger tracking and heatmaps enable 83.2% ROC-AUC suicide risk video analysis","Activity recognition and platform segmentation reach 83.2% ROC-AUC for metro SRA","Interpretable pipeline benchmarks metro suicide risk at 83.2% ROC-AUC"]},"model":"grok-4.3","cost_usd":0.006819,"raw_usage":{"total_tokens":3158,"prompt_tokens":646,"num_sources_used":0,"completion_tokens":92,"cost_in_usd_ticks":68187000,"prompt_tokens_details":{"text_tokens":646,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2420,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":646,"tokens_out":92,"duration_ms":27819,"temperature":1.0,"reasoning_tokens":2420,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T16:41:34.890151+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A controlled test on new metro surveillance footage in which the framework produces risk scores that do not correlate with actual recorded incidents or intervention outcomes at rates above random chance.","supporting_citations":[],"review_version":2}