{"id":"ae08142c-6ea3-47c2-b78e-2e5a85267940","arxiv_id":"2608.12548","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Neurotypical adults show increased movement variability during socially-framed dance imitation, while autistic adults remain consistent, and this contrast (SCSI) yields 79.2% balanced accuracy in group classification.","lead":"This paper analyzes 3D motion capture data from autistic and neurotypical adults doing dance imitation, and finds that neurotypical adults vary their movements more when the imitation is socially framed while autistic adults stay consistent. The authors propose a Social Context Sensitivity Index and report 79.2% balanced accuracy in classifying the two groups.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"SCSI's construct validity rests on unverified stimulus matching: the duo animation may differ in complexity, not just social framing.","rationale":"The reader's weakest_assumption correctly identifies the solo/duo stimulus contrast as the load-bearing condition for the paper's central claim. I agree: the paper internally justifies the contrast with a single sentence and provides no stimulus specifications, matching statistics, or manipulation checks. The observed reaction-time trend in Table I (both groups slower and more variable in duo) is suggestive but not decisive; it makes the task-difficulty confound concrete. Other weaknesses (small n, unreported exclusions, uncorrected multiple comparisons) are real but secondary; they would be addressable without changing the interpretation. The decisive issue is construct validity: if the duo stimulus is more complex, the SCSI is not a measure of social adaptation, and the proposed biomarker loses its specificity. I would therefore keep the CONDITIONAL verdict and add the stimulus-matching test as a required revision. The paper has merit: the DTW pipeline is interpretable, the progressive feature analysis is sensible, and the classification scheme uses nested CV; those strengths do not, however, resolve the stimulus-confound concern.","tokens_in":12028,"tokens_out":4269,"duration_ms":49152,"concrete_test":"Run a three-condition experiment with the same participants: (1) solo point-light dancer; (2) socially-framed duo animation as in the paper; (3) non-social control animation with two point-light figures performing independent, non-interacting movements matched to the duo stimulus for number of actors, joint speed, and spatial dispersion. Compute SCSI between solo and each companion condition. If SCSI in condition 3 approaches the effect in condition 2, the contrast tracks motor complexity rather than social framing; if it is near zero, the social-framing interpretation is supported. Report a manipulation check of perceived social presence for each animation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the SCSI measures 'social context sensitivity' (Section V) depends on the claim in Section I that solo and duo point-light animations differ only in social framing 'without changing the fundamental motor demands of the task.' The Methods (Section III-A) provide no stimulus details: content of the two animations, number of actors, interaction between actors, movement speed, or visual complexity. No manipulation check is reported. The descriptive increase in reaction time for both groups in the duo condition (Table I: clinical 311 to 378 ms, control 321 to 386 ms) is consistent with the duo stimulus being more demanding or attentionally engaging for non-social reasons. If the duo condition is more complex, the group difference in SCSI (upper d=-1.107, lower d=-1.238) could reflect differential response to task difficulty rather than social adaptation, and the classification result (79.2% balanced accuracy) would inherit this confound. Without stimulus validation, the biomarker interpretation is underdetermined.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a computational pipeline for distinguishing autistic from neurotypical adults using 3D motion capture during dance imitation. The authors compute intra-participant Dynamic Time Warping (DTW) consistency scores for solo and socially-framed duo dance conditions, introduce a Social Context Sensitivity Index (SCSI) defined as the difference between duo and solo DTW consistency, and use these features in an SVM classifier with leave-one-subject-out cross-validation. They report that neurotypical adults increase movement variability in the duo condition, especially in upper and lower limbs, whereas autistic adults maintain consistency, with large SCSI effect sizes (d = -1.107 and -1.238) and a balanced classification accuracy of 79.2% for the combined feature set. The authors conclude that social context sensitivity in motor imitation is a robust biomarker of autism-related motor behavior and discuss implications for human-centric and assistive technologies.","tokens_in":1892,"tokens_out":1826,"duration_ms":53045,"significance":"If the central claims hold, the paper would offer a relatively interpretable, low-cost kinematic biomarker for autism that goes beyond single-condition motor assessments and connects social processing to motor variability. The use of DTW-based trial-to-trial consistency is sensible and the pipeline is described in enough detail to be reproduced, including LOSO cross-validation with nested hyperparameter tuning and within-fold standardization. The paper also points to a public dataset (Move4AS), which is a strength for future verification. However, the significance is currently constrained by three load-bearing concerns: the statistical evidence is marginal and uncorrected for multiple comparisons, the participant count in the analyses does not match the reported cohort sizes, and the SCSI is a deterministic function of the two DTW features used in the classifier, so the claim that the social contrast is the 'strongest discriminative signature' is not directly supported by the classification comparison.","major_comments":[{"comment":"The main statistical claims rest on uncorrected multiple comparisons. Across Table II, Table III, and the SCSI comparisons in Section IV-D, at least 12 Mann-Whitney tests are reported with no multiple-comparison correction. The upper-body duo difference has p = 0.044, and the global duo difference is not significant (p = 0.066). With a Bonferroni correction for 12 tests, the upper-body p-value would not survive, and the SCSI upper/lower p-values (0.007 and 0.015) would be weakened. Please report the total number of tests, apply an appropriate correction (e.g., Benjamini-Hochberg), and interpret borderline effects accordingly. The current wording in Sections IV-D and V, which describes the effects as 'robust' and 'significantly lower,' overstates what the uncorrected p-values support.","section":"IV-C and IV-D"},{"comment":"The participant numbers are inconsistent. Section III-A1 states 20 neurotypical and 14 autistic participants, but all analyses use n = 18 and n = 12, and Table IV references n = 30. No exclusion criteria, missing-data handling, or attrition is described anywhere in the Methods. This mismatch affects every reported p-value, effect size, and classification result. The authors must specify exactly which participants were excluded and why, and confirm that the reported statistics are based on the full usable sample. Without this clarification, the validity of the results cannot be assessed.","section":"III-A and Tables I-IV"},{"comment":"The claim that the SCSI is the 'strongest discriminative signature' (Section V) is not supported by the classifier comparison. In the 'Solo+Duo' scenario, the feature set includes DTW_solo, DTW_duo, and SCSI = DTW_duo - DTW_solo per joint group. Since SCSI is a deterministic linear combination of the two DTW features, the SVM decision function on (solo, duo, duo-solo) is no more expressive than the same decision function on (solo, duo). Adding SCSI cannot therefore provide independent information, and the improved balanced accuracy over 'Duo only' cannot be attributed specifically to the social contrast. Please perform an ablation that isolates the contrast, for example by comparing a feature set of {DTW_solo, DTW_duo} with one of {DTW_solo, DTW_duo, SCSI}, and also report feature-importance or permutation-based measures that directly evaluate the contribution of the SCSI. Alternatively, reframe the classification result as showing that a combination of both conditions performs better than either alone.","section":"III-F and Eq. (4)"},{"comment":"The central construct validity of the SCSI depends on the assumption that the solo and duo point-light animations differ only in social framing 'without changing the fundamental motor demands of the task' (Section I). No stimulus details are provided: the content of the animations, number of actors, movement speed, visual complexity, or interaction between actors. No manipulation check is reported. The descriptive increase in reaction time for both groups in the duo condition (Table I: clinical 311 to 378 ms, control 321 to 386 ms) is consistent with the duo stimulus being more complex or attentionally demanding for reasons unrelated to social cognition. The authors should provide stimulus properties, perceptual validation of matched motor demands, or at minimum a manipulation check of perceived social presence and difficulty. Without this, the SCSI may measure differential response to task difficulty rather than social adaptation, and the biomarker interpretation is underdetermined.","section":"I, III-A, and Table I"},{"comment":"The classification result of 79.2% balanced accuracy is presented without any uncertainty quantification or significance test. With n = 30 and LOSO cross-validation, the estimate has substantial variance, and it is not clear whether this performance is significantly above chance for this sample. Please report confidence intervals for balanced accuracy (e.g., via bootstrap over subjects) and a permutation-based p-value for the classifier against the null of no group difference. This is needed to support the strength of the biomarker claim.","section":"IV-E"}],"minor_comments":[{"comment":"The scatterplot uses one point per participant, but with n = 12 and n = 18 points are heavily overlapped; consider adding horizontal jitter or a violin/boxplot overlay so the reader can see the distributional shape and the bootstrap confidence intervals.","section":"Fig. 2"},{"comment":"In Table III, the core body duo condition shows p = 0.363 but d = -0.588, which is an unusual pairing of a nonsignificant p-value with a moderate effect size; please report the bootstrap confidence interval for this and other effect sizes so the reader can appreciate the uncertainty.","section":"IV-C"},{"comment":"Please include demographic details (age and sex) for the analyzed subsample after any exclusions, not only for the original cohort, since the reported means and sex ratios in Section III-A1 may not describe the participants in the statistical analyses.","section":"III-A1"},{"comment":"The reaction time threshold tau = mu_base + 3 sigma_base uses a participant-specific baseline; please state whether the baseline window was the same for all trials and report how many trials were excluded by the visual screening or by the reaction-time detection procedure.","section":"III-C"},{"comment":"The conclusion calls the SCSI a 'robust biomarker,' but the paper contains no replication or external validation; consider softening this claim in the abstract and conclusion until confirmatory evidence is available.","section":"VI"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely and potentially useful application, and the pipeline is clearly presented, but the load-bearing issues are substantial: the statistical evidence does not survive correction, the sample size discrepancy is unexplained, and the classifier experiment cannot isolate the SCSI contribution. These are fixable with reanalysis and additional reporting, but the manuscript is not ready in its current form. I would encourage the editor to request the authors provide the analysis code and the excluded-participant details during the revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this paper introduces a genuinely new feature—the Social Context Sensitivity Index, the solo-to-duo DTW contrast—and reports an interesting dissociation: neurotypical adults increase movement variability when imitating a socially-framed point-light dance, while autistic adults keep their movement consistency unchanged. If that holds, it is a plausible kinematic marker for social motor adaptation. The pipeline is clearly described and the progressive design (body shake, solo, duo) is thoughtful.\n\nWhat the paper does well: the SCSI itself is a simple, interpretable construct, and the group differences in SCSI are large (Cohen's d around -1.1 for upper and lower limbs). The authors are also careful to check reaction time and to compute DTW over the last three seconds to avoid onset artifacts.\n\nThe soft spots are real. First, the participant numbers don't add up: Methods say 20 controls and 14 autistic, but Tables I–III and the classification report n=18 and 12, with no explanation anywhere. That is a reporting gap that needs to be fixed or justified. Second, the statistical claims rely on uncorrected Mann-Whitney tests. Nine tests are reported across Table III and the SCSI analysis; the global duo effect is p=0.066 and the upper-body duo is p=0.044, which would not survive a simple Bonferroni correction. The SCSI tests (p=0.007 and 0.015) are stronger, but still need correction or pre-registration. Third, the construct validity of the SCSI rests on the assumption that solo and duo animations differ only in social framing. The Methods give no stimulus details and no manipulation check; the descriptive rise in reaction time in the duo condition for both groups suggests the duo stimulus may be more demanding or complex for non-social reasons. Without stimulus validation, the biomarker interpretation is underdetermined.\n\nThe most technical concern: SCSI is defined as DTW(duo) - DTW(solo), and the classifier in the \"Solo+Duo\" scenario includes SCSI alongside the two DTW features. For a linear SVM, SCSI is an affine linear combination of the other two features, so it adds no new linear representational capacity. The reported jump from 66.7% to 79.2% balanced accuracy when adding a redundant feature is suspicious; it could reflect overfitting in the nested CV or an artifact of feature standardization, and it should be explained or the comparison re-run without SCSI.\n\nOn balance, this is a worthwhile preliminary study with a novel feature, but the central claim that the contrast is a robust biomarker is not yet supported. I would send it to peer review, but ask for major revision: report exclusions, apply multiple-comparison correction, provide stimulus details or a manipulation check, and address the feature redundancy in classification.","headline":"A promising but under-validated biomarker: the new SCSI feature is interesting, but the statistical and construct-validity issues are substantial.","tokens_in":12759,"tokens_out":3525,"would_cite":false,"duration_ms":36106,"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":"Solo-to-social dance shift distinguishes autism at 79.2%","keywords":["autism","dance imitation","social context sensitivity","motion analysis","dynamic time warping","motor variability","classification"],"falsifier":"Show the identical duo animation once with social framing and once with non-social instructions; if the SCSI group difference persists, the result is driven by stimulus properties rather than social context.","tokens_in":11825,"feed_emoji":"🕺","tokens_out":16869,"duration_ms":120622,"temperature":0.7,"pith_summary":"The paper proposes that the way adults alter their movement consistency when imitation becomes socially framed is a reliable motor signature of autism. Using Dynamic Time Warping on 3D motion-capture data from solo and socially framed duo dance imitation, the authors define a per-participant Social Context Sensitivity Index (SCSI) as the difference in trial-to-trial consistency between the two contexts. They find that neurotypical adults increase movement variability in the duo condition, whereas autistic adults keep consistency stable, with large effect sizes in the limbs. A classifier using these features distinguishes autistic from neurotypical adults at 79.2% balanced accuracy under leave-one-subject-out cross-validation. The authors argue that this social-context contrast, not performance in either isolated condition, is the strongest discriminative signal, with implications for objective assessment and inclusive human-machine systems.","feed_headline":"Solo-to-social dance shift distinguishes autism at 79.2%","feed_subtitle":"The gap between solo and social imitation may serve as an autism biomarker.","key_machinery":"The central object is the Social Context Sensitivity Index (SCSI), a per-participant measure defined as the difference in mean pairwise Dynamic Time Warping (DTW) distance between the socially framed duo imitation condition and the solo imitation condition: $\\text{SCSI} = \\text{DTW}_{\\text{duo}} - \\text{DTW}_{\\text{solo}}$. DTW consistency itself is computed as the average DTW distance across all unique pairs of a participant's trials within a condition, using the final three seconds of movement, so lower values indicate more stereotyped repetition. The SCSI isolates the within-subject shift in movement variability caused by social framing: a positive value means the participant varied more when the stimulus was social. In this paper the SCSI is the feature that carries the discriminative signal, and it is combined with per-joint-group DTW consistency scores and fed into an SVM classifier with leave-one-subject-out cross-validation.","core_discovery":"The central finding is that autism-related motor differences appear in the modulation of movement variability by social framing, not in baseline motor ability or solitary imitation performance. In a motion-capture dance-imitation experiment, neurotypical adults produced significantly more variable movement across trials when the imitation stimulus was socially framed, while autistic adults produced nearly identical consistency in solo and duo conditions. The Social Context Sensitivity Index, defined as the difference between duo and solo DTW consistency, showed large group differences in upper (d = -1.107) and lower (d = -1.238) limbs, with autistic participants hovering near zero. The authors interpret this as attenuated top-down modulation of motor execution in response to socially embedded biological motion in autism. A classifier combining DTW consistency and SCSI reached 79.2% balanced accuracy, 75% sensitivity, and 83.3% specificity, and the authors conclude that the dynamic shift between solitary and social contexts is the strongest discriminative signature.","pith_inferences":["We infer that the SCSI's sensitivity to social framing may depend on the stimulus being truly matched in motor demands; a control study that keeps the animation identical and only varies social instructions would separate social adaptation from task difficulty.","The per-joint-group SCSI values shown in the paper could be used to explore individual differences in which body segments carry social adaptation, potentially revealing subtypes within the autistic group that group-level statistics obscure.","If the solo-versus-social contrast works for point-light stimuli, the same logic could be extended to other modalities (e.g., a live partner or a more realistic avatar) and to other populations with social-cognition differences, as a transdiagnostic motor marker."],"forward_implications":["Motor assessments that include only solitary, non-social actions may miss the autism-related signal; the contrast between social and non-social contexts is what carries the discriminative information.","The SCSI, as a per-participant measure, could serve as the basis for objective screening tools, though larger-sample validation is needed.","The observation that autistic adults maintain stable movement across contexts suggests that interventions might aim to train flexible motor modulation rather than rote repetition.","Interactive systems such as virtual agents and assistive robots should account for neurodivergent motor profiles, since social framing changes motor output differently across groups."],"supporting_citations":[{"why":"provides the multimodal motion capture dataset and preprocessing pipeline used for all analyses.","marker":"[16]"},{"why":"supplies the Dynamic Time Warping distance measure that forms the basis of the consistency scores.","marker":"[17]"},{"why":"shows DTW applied to motion capture data, supporting its use for within-participant consistency.","marker":"[31]"},{"why":"demonstrates that DTW captures intra-individual motor noise in autism, validating the consistency metric.","marker":"[32]"},{"why":"documents atypical biological motion processing in autism, which motivates the social framing hypothesis.","marker":"[11]"},{"why":"provides the model of social top-down response modulation used to interpret neurotypical variability in the duo condition.","marker":"[34]"},{"why":"offers a prior machine-learning approach to autism diagnosis from imitation kinematics, serving as a performance benchmark.","marker":"[2]"}],"fun_headline_variants":["Autism's motor signature: no solo-to-social dance shift","Dance imitation: social context gap flags autism at 79%","Neurotypicals vary, autism stays consistent in social dance","Attenuated social modulation of movement marks autism","Solo-to-social dance variability: autism biomarker at 79.2%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion stands on the assumption that the solo and duo point-light animations are identical in movement content and differ only in social framing, so the observed contrast is attributed to social context rather than to differences in complexity, attention, or timing.","fun_headline_variants_meta":{"raw":{"variants":["Autism's motor signature: no solo-to-social dance shift","Dance imitation: social context gap flags autism at 79%","Neurotypicals vary, autism stays consistent in social dance","Attenuated social modulation of movement marks autism","Solo-to-social dance variability: autism biomarker at 79.2%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000307,"raw_usage":{"total_tokens":1746,"prompt_tokens":920,"completion_tokens":826,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":536,"completion_tokens_details":{"reasoning_tokens":739}},"tokens_in":536,"tokens_out":826,"duration_ms":7414,"temperature":1.0,"reasoning_tokens":739,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:04:54.046766+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Show the identical duo animation once with social framing and once with non-social instructions; if the SCSI group difference persists, the result is driven by stimulus properties rather than social context.","supporting_citations":[{"cited_title":"A multimodal dataset addressing motor function in autism,","cited_arxiv_id":null,"evidence_quote":"provides the multimodal motion capture dataset and preprocessing pipeline used for all analyses."},{"cited_title":"A novel distance measure based on dynamic time warping to improve time series classification,","cited_arxiv_id":null,"evidence_quote":"supplies the Dynamic Time Warping distance measure that forms the basis of the consistency scores."},{"cited_title":"Dynamic time warping in classification and selection of motion capture data,","cited_arxiv_id":null,"evidence_quote":"shows DTW applied to motion capture data, supporting its use for within-participant consistency."},{"cited_title":"Enhanced motor noise in an autism subtype with poor motor skills,","cited_arxiv_id":null,"evidence_quote":"demonstrates that DTW captures intra-individual motor noise in autism, validating the consistency metric."},{"cited_title":"Recognizing biological motion and emotions from point-light displays in autism spectrum disorders,","cited_arxiv_id":null,"evidence_quote":"documents atypical biological motion processing in autism, which motivates the social framing hypothesis."},{"cited_title":"Social top-down response modulation (storm): a model of the control of mimicry in social interaction,","cited_arxiv_id":null,"evidence_quote":"provides the model of social top-down response modulation used to interpret neurotypical variability in the duo condition."},{"cited_title":"Applying ma- chine learning to kinematic and eye movement features of a movement imitation task to predict autism diagnosis,","cited_arxiv_id":null,"evidence_quote":"offers a prior machine-learning approach to autism diagnosis from imitation kinematics, serving as a performance benchmark."}],"review_version":1}