{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"invariant_mode_decomposition_design__psychology","round_index":0,"assessments":[{"hypothesis_id":"H1","search_queries":["personalized CBT ecological momentary assessment dynamic network eigenvalue unstable mode module selection","idiographic dynamic network psychotherapy treatment selection within person symptom transition","just-in-time adaptive intervention mental health dynamical systems symptom network treatment recommendation"],"sources":[{"source_id":"H1-S1","title":"Network Control Theory in Personalized Intervention Selection: A Proof-of-Concept Approach to Simulating Therapeutic Intervention Effects in Clinical Time-Series Data","publisher":"Taylor & Francis / Psychotherapy Research","url":"https://doi.org/10.1080/10503307.2026.2666624","source_class":"PRIMARY_RESEARCH","claims_supported":["Personalized intervention effects have been simulated from clinical time-series data using network control theory.","The underlying EMA data came from studies personalizing CBT with idiographic dynamic network models.","The work already links dynamic networks to therapeutic intervention selection."]},{"source_id":"H1-S2","title":"Clinical trial protocol NCT06517589","publisher":"ClinicalTrials.gov","url":"https://cdn.clinicaltrials.gov/large-docs/89/NCT06517589/Prot_001.pdf","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["The protocol constructs idiographic dynamic networks from EMA data.","It identifies key maladaptive processes and matches interventions to empirically supported change processes.","Personalized process-based therapy is being compared with routine CBT."]},{"source_id":"H1-S3","title":"A control theoretic approach to evaluate and inform ecological momentary interventions","publisher":"PubMed / U.S. National Library of Medicine","url":"https://pubmed.ncbi.nlm.nih.gov/39436927/","source_class":"PRIMARY_RESEARCH","claims_supported":["The study proposes personalized, data-driven intervention-delivery strategies using network control theory.","EMA supplies dynamic information about an individual's mental-health state.","Control-theoretic intervention selection is already an explicit computational-psychiatry approach."]}],"closest_analogue":"Network Control Theory in Personalized Intervention Selection, which uses idiographic clinical time series to simulate and rank personalized therapeutic interventions.","overlap":"Both estimate coupled within-person dynamics from intensive assessments, represent symptoms and processes as a networked state, simulate intervention effects, and use the model to select a treatment focus rather than simply targeting the largest symptom.","remaining_difference":"The located analogue does not clearly select a CBT module by first partitioning eigenmodes into stable and unstable classes and then ranking unstable modes by functional-outcome sensitivity. The testable distinction is whether this eigenmode-and-stability decision rule produces module choices or outcomes that differ from existing node/process-level network-control selection on the same patient time series.","classification":"POSSIBLE_DISTINCTION","disposition":"ADVANCE","rationale":"The treatment-selection function is already close, but the shallow search did not find the specific unstable-eigenmode decision rule. That bounded algorithmic and experimental difference warrants further review rather than a novelty conclusion."},{"hypothesis_id":"H2","search_queries":["adaptive exposure therapy difficulty physiological fear safety behavior algorithm trial by trial","closed-loop exposure therapy real time arousal adaptive difficulty anxiety","exposure therapy dynamical systems oscillation avoidance mode stability pacing","personalized automated exposure hierarchy adaptive virtual reality exposure therapy"],"sources":[{"source_id":"H2-S1","title":"Automated Personalized Exposure Therapy Based on Physiological Measures Using Experience-Driven Procedural Content Generation","publisher":"AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment","url":"https://ojs.aaai.org/index.php/AIIDE/article/view/18914","source_class":"PRIMARY_RESEARCH","claims_supported":["The proposed VRET framework predicts patient experience from physiological sensors.","It automatically adapts exposure parameters in response to the patient's physiology.","The stated motivation is to replace subjective or fixed personalization rules."]},{"source_id":"H2-S2","title":"Personalizing Exposure Therapy via Reinforcement Learning","publisher":"arXiv","url":"https://arxiv.org/abs/2504.14095","source_class":"PRIMARY_RESEARCH","claims_supported":["The system automatically adapts arachnophobia exposure content using physiological measures and reinforcement learning.","A human-subject study reports better performance than a rules-based personalization method.","The framework adjusts virtual stimuli to match a desired individual anxiety response."]},{"source_id":"H2-S3","title":"An adaptation algorithm for personalised virtual reality exposure therapy","publisher":"Elsevier / Computer Methods and Programs in Biomedicine","url":"https://www.sciencedirect.com/science/article/pii/S0169260722004588","source_class":"PRIMARY_RESEARCH","claims_supported":["The algorithm suggests individualized virtual-environment configurations for exposure therapy.","It uses a Rescorla-Wagner learning model to choose configurations intended to maximize learning.","A proof of concept compared algorithmic suggestions with psychotherapists' configurations."]}],"closest_analogue":"Physiology-driven reinforcement-learning VRET that automatically adjusts exposure content to an individual's response after or during exposure.","overlap":"Both use repeated exposure responses as feedback, personalize the next exposure's difficulty or content, operate as a closed-loop pacing system, and aim to improve learning while avoiding poorly calibrated escalation.","remaining_difference":"Located systems chiefly optimize a scalar target such as desired anxiety or predicted associative learning. The testable difference is whether a jointly estimated trial-to-trial mode over fear, approach, avoidance, arousal, and safety behavior—and its stability class—changes pacing decisions and safety-behavior outcomes relative to those scalar or reinforcement-learning controllers.","classification":"POSSIBLE_DISTINCTION","disposition":"ADVANCE","rationale":"Adaptive exposure pacing is established, but the searched analogues did not visibly use multivariate modal stability to distinguish harmful avoidance dynamics from harmless persistent arousal. The difference is specific enough for deeper comparison."},{"hypothesis_id":"H3","search_queries":["group psychotherapy social network analysis eigenvector centrality facilitator intervention","group therapy member influence network centrality participation facilitator targeting","psychotherapy group network analysis central members therapeutic factors","real-time group facilitation sociometric network centrality intervention"],"sources":[{"source_id":"H3-S1","title":"Social network diagnostics: a tool for monitoring group interventions","publisher":"BMC Public Health / PubMed Central","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3851809/","source_class":"PRIMARY_RESEARCH","claims_supported":["The tool measures group social networks during an intervention and generates data-driven action reports for group leaders.","Recommended actions include connecting particular members and disrupting unhelpful subgroups.","Leaders were instructed to implement network-derived recommendations in later sessions."]},{"source_id":"H3-S2","title":"Eigenvector centrality defines hierarchy and predicts graduation in therapeutic community units","publisher":"PubMed / U.S. National Library of Medicine","url":"https://pubmed.ncbi.nlm.nih.gov/34914758/","source_class":"PRIMARY_RESEARCH","claims_supported":["Eigenvector centrality was calculated from peer-correction networks in therapeutic communities.","Centrality represented members' positions in the therapeutic-community hierarchy.","Network position was tested as a predictor of treatment graduation."]},{"source_id":"H3-S3","title":"Strategic players for identifying optimal social network intervention subjects","publisher":"Elsevier / Social Networks","url":"https://www.sciencedirect.com/science/article/pii/S0378873316304695","source_class":"PRIMARY_RESEARCH","claims_supported":["The method uses network position to identify influential behavioral leaders for intervention.","It explicitly aims to select network members capable of diffusing behavior to receptive targets.","The work treats node targeting as an optimization problem rather than relying on local prominence alone."]}],"closest_analogue":"Social Network Diagnostics, which measures a behavioral-intervention group's network and gives its leader member-specific facilitation actions during the intervention.","overlap":"The analogue already replaces surface participation measures with network structure, identifies consequential members and subgroups, and translates the analysis into leader actions such as connecting or separating particular participants. Separate therapeutic-community research already applies eigenvector centrality to treatment-group member networks.","remaining_difference":"The stated experiment would estimate directed influence from psychotherapy-session behavior, use the dominant eigenvector specifically, and compare real-time prompting with speaking-time targeting. Those are narrower implementation and validation choices rather than a clearly different intervention concept.","classification":"OBVIOUS_COLLISION","disposition":"REJECT","rationale":"Network-derived, member-specific facilitation of behavioral groups is already demonstrated, and eigenvector centrality has already been applied in a therapeutic-group setting. The remaining difference is too narrow for advancement in this shallow screen."},{"hypothesis_id":"H4","search_queries":["treatment selection questionnaire item reduction heterogeneous treatment effect mental health assessment","precision treatment rules depression baseline moderators reduced questionnaire battery","predictive treatment selection lasso item reduction psychotherapy outcome assessment","differential treatment response feature selection mental health intake randomized trial"],"sources":[{"source_id":"H4-S1","title":"Analysis of Features Selected by a Deep Learning Model for Differential Treatment Selection in Depression","publisher":"Frontiers in Artificial Intelligence / PubMed","url":"https://pubmed.ncbi.nlm.nih.gov/33733120/","source_class":"PRIMARY_RESEARCH","claims_supported":["Models predicted response to four depression treatments from clinical and demographic features.","The analysis identified features associated with differential treatment response and nonresponse.","The stated clinical aim was to optimize treatment using selected, interpretable patient features."]},{"source_id":"H4-S2","title":"A demonstration of a multi-method variable selection approach for treatment selection: Recommending cognitive-behavioral versus psychodynamic therapy for mild to moderate adult depression","publisher":"Taylor & Francis / Psychotherapy Research","url":"https://www.tandfonline.com/doi/abs/10.1080/10503307.2018.1563312","source_class":"PRIMARY_RESEARCH","claims_supported":["The study uses multiple variable-selection methods to identify patient characteristics associated with differential treatment response.","Its decision target is recommending cognitive-behavioral versus psychodynamic therapy.","It already couples intake-feature selection to comparative treatment choice."]},{"source_id":"H4-S3","title":"Initial evaluation of a personalized advantage index to determine which individuals may benefit from mindfulness-based cognitive therapy for suicide prevention","publisher":"PubMed / U.S. National Library of Medicine","url":"https://pubmed.ncbi.nlm.nih.gov/39306938/","source_class":"PRIMARY_RESEARCH","claims_supported":["A Personalized Advantage Index was built from 55 baseline demographic, clinical, and neurocognitive variables.","Feature selection and treatment-arm models were used to predict differential response between two treatment pathways.","The resulting model was intended to match patients to the treatment with greater predicted advantage."]}],"closest_analogue":"Multi-method variable selection for recommending CBT versus psychodynamic therapy from baseline patient characteristics.","overlap":"Both start with a broad intake feature set, select a smaller action-relevant subset based on differential response rather than variance alone, and use the retained information to choose between treatments. Personalized Advantage Index work supplies the same treatment-specific decision target and prospective baseline workflow.","remaining_difference":"The hypothesis packages the selected information as a battery of at most half the original items and uses SVD plus structured decision residuals with an explicit noninferiority margin. Those are assessment-engineering and validation details layered onto an established treatment-moderator feature-selection design.","classification":"OBVIOUS_COLLISION","disposition":"REJECT","rationale":"Outcome-guided feature reduction for differential mental-health treatment selection is already direct prior work. SVD, residual auditing, and a half-length constraint do not create a sufficient conceptual separation at this screening stage."},{"hypothesis_id":"H5","search_queries":["psychiatric relapse early warning dynamical systems critical slowing down symptom network eigenvalue","depression relapse personalized early warning signals ecological momentary assessment network connectivity","mode rotation spectral gap mental health relapse monitoring","critical transition early warning psychiatric symptoms autocorrelation variance relapse"],"sources":[{"source_id":"H5-S1","title":"Critical slowing down as early warning for the onset and termination of depression","publisher":"Proceedings of the National Academy of Sciences / PubMed Central","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3890822/","source_class":"PRIMARY_RESEARCH","claims_supported":["The study proposes smartphone mood monitoring for early warning of transitions into and out of depression.","It frames depression as a coupled dynamical system with alternative stable states.","The mathematical account explicitly relates critical slowing to the dominant eigenvalue approaching a stability boundary."]},{"source_id":"H5-S2","title":"Anticipating the direction of symptom progression using critical slowing down: a proof-of-concept study","publisher":"Epidemiology and Psychiatric Sciences / PubMed Central","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8781362/","source_class":"PRIMARY_RESEARCH","claims_supported":["The study tests whether early slowing occurs along a specific combination of symptom variables.","It extends generic warning signals to predict the direction of symptom worsening or improvement.","The approach therefore already uses a direction in multivariate symptom state space as a relapse-relevant warning."]},{"source_id":"H5-S3","title":"Personalized relapse prediction in patients with major depressive disorder using digital biomarkers","publisher":"Scientific Reports / PubMed Central","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10616277/","source_class":"PRIMARY_RESEARCH","claims_supported":["The study develops an N-of-1 multivariate anomaly-detection framework for near-term depression relapse.","Patient-specific deviations in actigraphy trigger active symptom assessment.","The intended workflow is early clinical intervention before an impending relapse."]},{"source_id":"H5-S4","title":"Using Network Theory to Predict Depression Onset and Build a Personalized Early Warning System","publisher":"CORDIS, European Commission","url":"https://cordis.europa.eu/project/id/949059/results","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["WARN-D was an ERC-funded project explicitly devoted to a personalized network-theory early-warning system for depression.","Its listed outputs include EMA methods and early-warning-system research.","The project establishes that personalized network-based warning for depression is an active, organized research program."]}],"closest_analogue":"Personalized critical-slowing-down monitoring, where multivariate momentary-state dynamics and the dominant eigenvalue or its directional mode warn of an approaching depressive transition.","overlap":"Both monitor coupled within-person indicators instead of aggregate symptom totals, use rolling dynamical changes as pre-relapse signals, interpret weakening stability in spectral terms, and aim to trigger earlier clinical intervention.","remaining_difference":"The hypothesis substitutes or supplements established autocorrelation, variance, connectivity, dominant-eigenvalue, and anomaly signals with retained-mode rotation, spectral-gap collapse, and structured residual growth. The alert workflow and causal purpose remain the same; only the monitored spectral diagnostics differ.","classification":"OBVIOUS_COLLISION","disposition":"REJECT","rationale":"Personalized spectral or dynamical early warning of depressive relapse is already explicit, including dominant-eigenvalue theory and directional multivariate modes. Gap, rotation, and residual thresholds appear to be metric variants within that established design."}],"nominated_ids":["H1","H2"],"replenishment_recommended":false}