VAST · GPT-5.6-Terra · C6++

workflow vastopolis-outbreak-response-revised template: loop(define-unit → build-model → characterise → visualise → assess → build-model) → characterise → define-unit → characterise → contextualise → visualise → abstract → generate-knowledge → generate-knowledge → characterise → build-model → loop(characterise → contextualise → visualise → abstract → characterise → assess → build-model) → visualise → generate-knowledge → assess → characterise → generate-knowledge → generate-knowledge → loop(define-unit → characterise → visualise → assess → generate-knowledge) description: "Revised human-in-the-loop visual analytics workflow for identifying credible flu-like illness signals, locating candidate outbreak origins, comparing transmission hypotheses, validating findings with clinical and environmental evidence, and adaptively prioritising emergency response in Vastopolis."

artifact D_posts : entities origin: given internal structure: elementary embedment: {space, time} features: - id: timestamp value structure: atomic value type: temporal description: "Time at which the message was posted" - id: gps_location value structure: atomic value type: spatial description: "Mobile-device GPS coordinate" - id: text value structure: atomic value type: text description: "Microblog message text" - id: pseudonymous_account_id value structure: atomic value type: reference description: "Access-controlled pseudonymous identifier, used only for duplicate and probable-automation detection" description: "Raw geolocated microblog posts collected during the incident" artifact D_city_features : entities origin: given internal structure: elementary embedment: space features: - id: feature_type value structure: atomic value type: categorical description: "Highway, hospital, landmark, water body, or related feature" - id: geometry value structure: atomic value type: spatial description: "Geographic geometry" - id: name value structure: atomic value type: text description: "Feature name where available" description: "Metropolitan map features" artifact D_zones : entities origin: given internal structure: region embedment: space features: - id: zone_name value structure: atomic value type: text description: "City-zone identifier" - id: population value structure: atomic value type: numeric description: "Resident population" description: "Thirteen city zones with population statistics" artifact D_weather : entities origin: given internal structure: elementary embedment: time features: - id: wind_speed value structure: atomic value type: numeric description: "Observed wind speed" - id: wind_direction value structure: atomic value type: numeric description: "Observed wind direction" description: "Available daily weather observations" artifact D_confirmation_data : entities origin: given internal structure: elementary embedment: {space, time} features: - id: evidence_type value structure: atomic value type: categorical description: "Confirmed case, laboratory result, emergency-department activity, death record, water sample, environmental sample, or other external evidence" - id: observation_time value structure: atomic value type: temporal description: "Clinical onset, test, visit, sampling, or record time" - id: approximate_location value structure: atomic value type: spatial description: "Privacy-preserving residence, exposure, service-area, or sampling location" - id: result value structure: atomic value type: categorical description: "Observed result or status" description: "External clinical and environmental confirmation data. This artifact may be empty during the initial run and populated later." artifact D_operational_updates : entities origin: given internal structure: elementary embedment: time features: - id: update_type value structure: atomic value type: categorical description: "New surveillance signal, field-investigation result, intervention record, resource deployment, or public communication" - id: update_time value structure: atomic value type: temporal description: "Time of update" - id: affected_area value structure: atomic value type: spatial description: "Affected area, where applicable" description: "Subsequent operational updates received while the emergency response is active" artifact K_epidemiology : knowledge(D_posts, D_weather, D_city_features) origin: given representation form: "domain assumptions and interpretation guidance" description: "Public-health expertise concerning symptom terminology, plausible reporting delay, expected transmission mechanisms, realistic lags, clinically meaningful severity indicators, and limitations of social-media observations" artifact K_roles : knowledge(D_posts, D_confirmation_data) origin: given representation form: "role assignments and decision authority" description: "The data analyst manages preprocessing and computational analysis; the epidemiologist interprets illness signals and mechanisms; the environmental-health or field investigator evaluates candidate sources and sampling; the emergency manager authorises deployment subject to capacity and equity constraints." artifact S_governance : specification origin: given representation form: "privacy, ethics, and equity constraints" description: "Use only authorised pseudonymous data; prohibit individual identification, contact tracing, or publication of identifiable posts; aggregate outputs where possible; retain uncertainty for low-posting areas; and prevent low platform participation from being interpreted as low need." artifact S_cleaning_initial : specification origin: given representation form: "filtering and data-quality rules" description: "Rules for duplicate removal, probable-bot handling, coordinate and timestamp validation, account-level privacy protection, and retention of data-quality flags" artifact S_content_initial : specification origin: given representation form: "text-classification configuration" description: "Initial symptom vocabulary, topic-model settings, report categories, confidence thresholds, and exclusion rules" artifact S_signal_initial : specification origin: given representation form: "signal-validity and aggregation criteria" description: "Spatial resolutions, one-, three-, and six-hour bins, population normalization, posting-volume adjustment, minimum independent reporting support, low-volume uncertainty rules, and criteria distinguishing a credible symptom signal from media or automation artifacts" artifact S_onset_uncertainty : specification origin: given representation form: "illness-time inference rules" description: "Rules for distinguishing posting time from stated symptom onset, extracting expressions such as 'since yesterday', assigning onset intervals rather than point estimates, and retaining reporting-delay uncertainty when onset cannot be inferred" artifact S_origin_detection : specification origin: given representation form: "cluster and origin-ranking criteria" description: "Space-time clustering parameters, sustained-onset definition, persistence requirements, independent-account support rules, candidate-origin ranking criteria, and sensitivity ranges" artifact S_hypothesis_preregistration : specification origin: given representation form: "hypothesis-testing and holdout protocol" description: "Rules requiring candidate-origin discovery on discovery data only; pre-specification of mechanism-specific expected patterns, lag ranges, predictors, comparison criteria, and temporally or spatially held-out validation units before final model evaluation" artifact S_bias_controls : specification origin: given representation form: "confounding and reporting-bias controls" description: "Require adjustment for population, platform activity, local posting coverage, hospital-related reporting, news amplification, spatial dependence, and uncertainty caused by sparse reporting" artifact S_model_initial : specification origin: given representation form: "initial model configuration" description: "Transparent count or rate model forms for person-to-person, airborne, water-related, point-source, and mixed explanations; candidate covariates; complexity penalty; and held-out evaluation rules" artifact S_evaluation : specification origin: given representation form: "quality and decision criteria" description: "Explicit criteria for acceptable content classification, credible symptom signals, cluster robustness, model generalisation, residual acceptability, hypothesis support labels, and circumstances in which conclusions must remain inconclusive" artifact S_sensitivity : specification origin: given representation form: "sensitivity-analysis design" description: "Required variation of classification thresholds, temporal-bin width, spatial-unit resolution, clustering settings, source definitions, lag choices, normalization methods, and model covariates; conclusions are labelled robust, sensitive, or unsupported" artifact S_resource_priority : specification origin: given representation form: "resource-prioritisation rules" description: "Prioritise areas using current symptom intensity, growth, severe-report signal, population exposure, healthcare access, uncertainty, vulnerability where available, equity constraints, and operational capacity; do not use social-media volume alone as a need score" artifact S_intervention_evaluation : specification origin: given representation form: "post-intervention assessment criteria" description: "Rules for comparing conditions before and after interventions while accounting for changes in posting volume, public attention, testing, and reporting practice; define triggers for a full analytical re-run" assign: S_cleaning_current := S_cleaning_initial S_content_current := S_content_initial S_signal_current := S_signal_initial S_model_current := S_model_initial

loop L_signal_preparation: purpose: "Create a privacy-preserving, semantically credible, and operationally usable illness-report signal before origin or mechanism inference begins" until: "The epidemiologist and data analyst judge that classified symptom reports and aggregated signals meet the explicit validity criteria, or judge that the available social-media data are not adequate for outbreak inference" body: transform T_clean_posts : intent: define-unit manner: "filter and quality-control" input: D_posts, S_cleaning_current, S_governance output: D_clean_posts actor: machine description: "Remove duplicates, likely automated posting, and invalid coordinates or timestamps while preserving data-quality flags and privacy restrictions" artifact D_clean_posts : entities internal structure: elementary embedment: {space, time} features: - id: text value structure: atomic value type: text description: "Retained message text" - id: location value structure: atomic value type: spatial description: "Validated location" - id: posting_time value structure: atomic value type: temporal description: "Validated post timestamp" - id: quality_flag value structure: atomic value type: categorical description: "Data-quality and uncertainty indicator" description: "Quality-controlled, privacy-constrained microblog posts" transform T_build_content_model : intent: build-model manner: "train text classifier and topic model" input: D_clean_posts, S_content_current, K_epidemiology output: M_content_model actor: hybrid description: "The data analyst configures computational text processing; the epidemiologist validates symptom concepts, ambiguous terms, and relevant exclusions" artifact M_content_model : model(D_clean_posts, K_epidemiology) model type: "text classifier and topic model" representation form: "keyword rules, topic structure, and confidence model" description: "Model for identifying likely flu-like symptom reports and distinguishing them from news, concern, misinformation, and unrelated content" transform T_classify_posts : intent: characterise manner: "classify and score message content" input: D_clean_posts, M_content_model, S_content_current output: F_content_classification actor: machine description: "Assign content category and confidence, including self-reported symptoms, reports about others, severe illness or death, healthcare seeking, news or concern, and unrelated messages" artifact F_content_classification : feature(D_clean_posts) value structure: vector value type: {categorical, numeric} representation form: "content labels and confidence values" description: "Probabilistic illness-related content classification" transform T_select_health_reports : intent: define-unit manner: "filter by confidence and report type" input: D_clean_posts, F_content_classification, S_content_current, S_governance output: D_symptom_reports, D_severe_reports actor: machine description: "Select high-confidence symptom reports and separately retain severe-illness, death, and healthcare-seeking reports" artifact D_symptom_reports : entities internal structure: elementary embedment: {space, time} features: - id: symptom_confidence value structure: atomic value type: numeric description: "Confidence that the post reports flu-like illness" - id: report_type value structure: atomic value type: categorical description: "Self-report or report about another person" description: "High-confidence flu-like symptom-report posts" artifact D_severe_reports : entities internal structure: elementary embedment: {space, time} features: - id: severity_type value structure: atomic value type: categorical description: "Severe illness, death, hospital, or healthcare-seeking category" description: "Severe-report signal retained separately from symptom reports" transform T_infer_illness_time : intent: characterise manner: "extract onset expressions and reporting-delay intervals" input: D_symptom_reports, S_onset_uncertainty, K_epidemiology output: F_illness_time actor: hybrid description: "Extract stated onset expressions when available and assign an uncertain illness-onset interval; otherwise retain posting time only as a delayed and uncertain proxy" artifact F_illness_time : feature(D_symptom_reports) value structure: vector value type: temporal representation form: "onset interval and reporting-delay uncertainty" description: "Estimated symptom-onset timing, or explicitly uncertain posting-time proxy, for each symptom report" transform T_construct_signal_units : intent: define-unit manner: "multi-resolution spatiotemporal aggregation" input: D_symptom_reports, D_severe_reports, F_illness_time, D_zones, S_signal_current output: D_analysis_areas, D_signal_units actor: machine description: "Create zone, fine-grid, or adaptive spatial units and aggregate reports into multiple time resolutions" artifact D_analysis_areas : entities internal structure: region embedment: space features: - id: area_type value structure: atomic value type: categorical description: "Zone, grid cell, or adaptive area" - id: population value structure: atomic value type: numeric description: "Population estimate" description: "Spatial areas used for outbreak and equity analysis" artifact D_signal_units : entities internal structure: episode embedment: {space, time} features: - id: temporal_extent value structure: atomic value type: temporal description: "Time-bin interval" - id: spatial_area value structure: atomic value type: reference description: "Associated analysis area" description: "Area-time units containing aggregated illness-report activity" transform T_compute_signal_and_bias_features : intent: characterise manner: "aggregate, normalise, and estimate reporting bias" input: D_signal_units, D_clean_posts, D_analysis_areas, D_city_features, S_signal_current, S_bias_controls output: F_outbreak_indicators, F_reporting_bias actor: machine description: "Compute symptom counts, population-normalised rates, symptom share of all local posts, growth, severe-report activity, uncertainty, hospital-related reporting, news amplification, local platform coverage, and spatial-dependence controls" artifact F_outbreak_indicators : feature(D_signal_units) value structure: vector value type: numeric representation form: "rates, counts, growth, severity, and uncertainty" description: "Outbreak indicators derived from social-media reports" artifact F_reporting_bias : feature(D_analysis_areas) value structure: vector value type: {numeric, categorical} representation form: "coverage and confounding indicators" description: "Indicators of platform-activity variation, sparse coverage, hospital-reporting effects, news amplification, and spatial dependence" transform T_arrange_signal_map : intent: contextualise manner: "map-based" input: D_signal_units, D_analysis_areas output: A_signal_map actor: machine description: "Arrange area-time signal units in their geographic areas" artifact A_signal_map : arrangement(D_signal_units) context: D_analysis_areas principle: "spatial-area membership" description: "Geographic arrangement of aggregated outbreak signals" transform T_visualise_preparation_evidence : intent: visualise manner: "linked text, map, and temporal quality-review display" input: D_clean_posts, F_content_classification, A_signal_map, F_outbreak_indicators, F_reporting_bias, D_severe_reports, D_city_features output: V_preparation_evidence actor: machine description: "Show sampled source posts, category summaries, animated maps, area time series, severe reports, total posting activity, and uncertainty indicators" artifact V_preparation_evidence : visualisation(D_clean_posts, F_content_classification, A_signal_map, F_outbreak_indicators, F_reporting_bias, D_severe_reports) layout: "linked document list, topic summary, geographic map, temporal heatmap, and area time-series views" form: "text excerpts, coloured areas, points, lines, and uncertainty bands" encoding: "Text panel exposes source messages; colour encodes content class or symptom rate; map position follows A_signal_map; uncertainty, total activity, and severe reports are shown as linked overlays" description: "Evidence display for assessing semantic validity, representativeness limitations, and outbreak-signal credibility" transform T_assess_signal_preparation : intent: assess manner: "evaluate classification, signal validity, and ethical adequacy" input: V_preparation_evidence, F_content_classification, F_outbreak_indicators, F_reporting_bias, S_evaluation, S_governance, K_epidemiology, K_roles output: K_signal_preparation_assessment, S_cleaning_refined, S_content_refined, S_signal_refined actor: human description: "The data analyst and epidemiologist assess whether patterns reflect plausible illness reports rather than automated activity, news, hospital reporting, or sparse-platform artifacts; they also verify privacy and equity constraints. They determine whether to refine the signal or halt inference." artifact K_signal_preparation_assessment : knowledge( V_preparation_evidence, F_content_classification, F_outbreak_indicators, F_reporting_bias) representation form: "quality judgment and proceed-or-halt decision" description: "Assessment of content credibility, signal adequacy, representativeness limits, privacy compliance, and whether origin and transmission inference may proceed" artifact S_cleaning_refined : specification representation form: "revised filtering and quality-control rules" description: "Updated duplicate, automation, location, timestamp, or privacy-handling rules when required" artifact S_content_refined : specification representation form: "revised text-classification configuration" description: "Updated vocabulary, topic interpretation, categories, confidence thresholds, and exclusions" artifact S_signal_refined : specification representation form: "revised aggregation and signal-validity criteria" description: "Updated spatial or temporal resolution, normalization, independence, uncertainty, or artifact-detection settings" if K_signal_preparation_assessment indicates refinement is needed: then: assign: S_cleaning_current := S_cleaning_refined S_content_current := S_content_refined S_signal_current := S_signal_refined else: exit loop L_signal_preparation end loop L_signal_preparation

transform T_partition_discovery_and_holdout : intent: define-unit manner: "temporal or spatial holdout partitioning" input: D_signal_units, S_hypothesis_preregistration output: D_discovery_units, D_holdout_units actor: machine description: "Partition signal units into discovery data for identifying candidate origins and a held-out subset for later model evaluation" artifact D_discovery_units : entities internal structure: episode embedment: {space, time} description: "Signal units reserved for exploratory origin discovery" artifact D_holdout_units : entities internal structure: episode embedment: {space, time} description: "Signal units reserved for testing pre-specified hypotheses"

transform T_cluster_early_outbreak_activity : intent: define-unit manner: "space-time density clustering" input: D_discovery_units, F_outbreak_indicators, F_illness_time, S_origin_detection output: D_outbreak_clusters, F_cluster_membership actor: machine description: "Identify persistent concentrations of early symptom reports, using onset uncertainty and independent-report support where possible" artifact D_outbreak_clusters : entities internal structure: group/cluster embedment: {space, time} features: - id: cluster_extent value structure: vector value type: {spatial, temporal} description: "Spatial and temporal cluster extent" description: "Persistent candidate outbreak clusters identified from discovery data" artifact F_cluster_membership : feature(D_discovery_units) value structure: atomic value type: reference description: "Outbreak-cluster membership of discovery signal units" transform T_characterise_clusters : intent: characterise manner: "compute onset, persistence, expansion, and robustness" input: D_outbreak_clusters, D_discovery_units, F_outbreak_indicators, F_reporting_bias, S_origin_detection output: F_cluster_descriptors actor: machine description: "Compute onset intervals, intensity, persistence, population-normalised severity, downstream temporal ordering, reporting-bias indicators, and parameter-sensitivity descriptors" artifact F_cluster_descriptors : feature(D_outbreak_clusters) value structure: vector value type: {numeric, temporal} representation form: "candidate-origin evidence descriptors" description: "Features used to compare candidate origin areas" transform T_arrange_clusters : intent: contextualise manner: "map-based" input: D_outbreak_clusters, D_analysis_areas output: A_cluster_map actor: machine description: "Place candidate outbreak clusters in metropolitan geography" artifact A_cluster_map : arrangement(D_outbreak_clusters) context: D_analysis_areas principle: "cluster centroid and spatial extent" description: "Map arrangement of early outbreak clusters" transform T_visualise_origin_evidence : intent: visualise manner: "animated onset and expansion map" input: A_cluster_map, F_cluster_descriptors, F_reporting_bias, D_city_features, F_outbreak_indicators output: V_origin_evidence actor: machine description: "Display candidate clusters, onset intervals, expansion contours, uncertainty, and nearby hospitals, highways, landmarks, and water bodies" artifact V_origin_evidence : visualisation(A_cluster_map, F_cluster_descriptors, F_reporting_bias, D_city_features, F_outbreak_indicators) layout: "time-enabled map with linked candidate ranking and time series" form: "coloured regions, contours, points, ranked bars, and uncertainty marks" encoding: "Colour represents estimated onset; size represents intensity; uncertainty reflects onset and coverage uncertainty; map overlays represent city features and reporting-bias indicators" description: "Visual evidence for interpreting early sustained outbreak activity" transform T_interpret_origin_candidates : intent: abstract manner: "rank persistent early clusters" input: V_origin_evidence, F_cluster_descriptors, K_signal_preparation_assessment, K_epidemiology output: P_origin_candidates actor: hybrid description: "The analyst identifies robust early clusters and distinguishes a possible infection source from the earlier appearance of a social-media or healthcare-seeking signal" artifact P_origin_candidates : pattern(D_outbreak_clusters, F_cluster_descriptors) representation form: "ranked candidate-origin patterns" description: "Candidate source areas with observed timing, expansion, robustness, reporting-bias caveats, and uncertainty" transform T_formulate_origin_assessment : intent: generate-knowledge manner: "formulate-statements" input: P_origin_candidates, V_origin_evidence, K_signal_preparation_assessment, K_epidemiology output: K_origin_assessment actor: human description: "The epidemiologist records a ranked, uncertainty-aware candidate-origin assessment, explicitly distinguishing observed signal origin from confirmed infection origin" artifact K_origin_assessment : knowledge(P_origin_candidates, V_origin_evidence) representation form: "ranked evidence statements" description: "Candidate outbreak-origin areas, supporting and contradictory observations, confidence, and reporting-bias caveats"

transform T_preregister_transmission_hypotheses : intent: generate-knowledge manner: "formulate-statements and methodological decisions" input: K_origin_assessment, K_epidemiology, D_city_features, D_weather, S_hypothesis_preregistration, S_evaluation output: K_transmission_hypotheses, S_hypothesis_models actor: human description: "Before evaluating held-out outcomes, the epidemiologist and data analyst state expected patterns, lags, predictors, contradictions, and comparison rules for person-to-person, airborne, water-related, common-point-source, reporting-artifact, and mixed explanations" artifact K_transmission_hypotheses : knowledge(K_origin_assessment, D_city_features, D_weather) representation form: "competing explanatory statements" description: "Pre-specified expected observable patterns and disconfirming evidence for candidate transmission mechanisms" artifact S_hypothesis_models : specification representation form: "pre-registered model and evaluation configuration" description: "Mechanism-specific predictors, candidate sources, lag ranges, discovery-versus-holdout roles, matched comparison areas, complexity rules, and decision criteria fixed before held-out evaluation"

transform T_compute_exposure_predictors : intent: characterise manner: "derive lagged spatial-exposure and bias-control features" input: D_signal_units, F_outbreak_indicators, F_reporting_bias, D_analysis_areas, D_city_features, D_weather, K_origin_assessment, S_hypothesis_models, S_bias_controls output: F_exposure_predictors actor: machine description: "Compute prior local rates, adjacent-area and highway-connected rates, candidate-source distance, downwind alignment, wind speed, water proximity, reporting-volume controls, hospital proximity, spatial dependence, and population-related factors" artifact F_exposure_predictors : feature(D_signal_units) value structure: vector value type: numeric representation form: "lagged exposure, connection, and confounding predictors" description: "Predictors used to compare competing transmission mechanisms" transform T_build_initial_transmission_models : intent: build-model manner: "fit transparent competing spatiotemporal models" input: D_discovery_units, F_outbreak_indicators, F_exposure_predictors, K_transmission_hypotheses, S_hypothesis_models, S_model_current, S_bias_controls output: M_transmission_models_initial actor: machine description: "Fit alternative contact-propagation, airborne, water-related, point-source, and penalised mixed-mechanism models using discovery data only" artifact M_transmission_models_initial : model(D_discovery_units, F_outbreak_indicators, F_exposure_predictors) model type: "suite of transparent spatiotemporal count or rate regression models" representation form: "mechanism-specific predictor sets with spatial-dependence and reporting-bias controls" description: "Competing explanatory and predictive models of outbreak activity" assign: M_transmission_models := M_transmission_models_initial loop L_hypothesis_evaluation: purpose: "Evaluate pre-specified transmission explanations, inspect residual structure, and refine models only when diagnostic and sensitivity evidence justifies a revision" until: "The models meet the stated generalisation and interpretability criteria, or the evidence remains too weak to distinguish mechanisms" body: transform T_apply_transmission_models : intent: characterise manner: "predict and compute held-out diagnostics" input: D_signal_units, D_holdout_units, F_outbreak_indicators, F_exposure_predictors, M_transmission_models, S_hypothesis_models output: F_model_outputs actor: machine description: "Produce predicted rates, uncertainty intervals, residuals, fit measures, and explicitly held-out temporal or spatial validation measures for every candidate model" artifact F_model_outputs : feature(D_signal_units) value structure: vector value type: numeric representation form: "observed, predicted, residual, and held-out performance values" description: "Comparative diagnostic outputs for all transmission models" transform T_arrange_model_diagnostics : intent: contextualise manner: "map-based" input: D_signal_units, D_analysis_areas output: A_model_map actor: machine description: "Arrange predictions and residuals by geographic analysis area" artifact A_model_map : arrangement(D_signal_units) context: D_analysis_areas principle: "spatial-area membership" description: "Geographic arrangement for model diagnostics" transform T_visualise_model_diagnostics : intent: visualise manner: "comparative residual and validation display" input: A_model_map, F_model_outputs, D_weather, D_city_features, K_transmission_hypotheses, F_reporting_bias output: V_model_diagnostics actor: machine description: "Show observed-versus-predicted activity, residual maps, uncertainty, held-out results, lag plots, wind vectors, upwind/downwind comparisons, water proximity, and coverage or reporting-bias context" artifact V_model_diagnostics : visualisation(A_model_map, F_model_outputs, D_weather, D_city_features, F_reporting_bias) layout: "linked maps, time series, observed-predicted plots, residual panels, lag plots, and comparative score charts" form: "coloured areas, lines, arrows, scatterplots, and uncertainty bands" encoding: "Map colour encodes residual or predicted rate; wind arrows encode direction and speed; time plots show lagged relationships; panels compare mechanisms and held-out performance" description: "Diagnostic environment for evaluating transmission explanations" transform T_interpret_model_patterns : intent: abstract manner: "perception-and-interpretation of explanatory patterns" input: V_model_diagnostics, F_model_outputs, K_transmission_hypotheses, K_epidemiology output: P_model_patterns actor: human description: "The epidemiologist identifies spatial propagation waves, stable wind-related lag patterns, water-aligned concentrations, point-source signatures, or coherent residual anomalies" artifact P_model_patterns : pattern(F_model_outputs, V_model_diagnostics) representation form: "annotated explanatory and residual patterns" description: "Observed model successes, failures, cross-impact patterns, and coherent unexplained residual structures" transform T_run_sensitivity_analysis : intent: characterise manner: "systematic multi-specification robustness analysis" input: D_clean_posts, F_content_classification, D_signal_units, F_outbreak_indicators, F_exposure_predictors, M_transmission_models, S_sensitivity, S_hypothesis_models output: F_sensitivity_results actor: machine description: "Re-evaluate material conclusions under required variations in classification threshold, spatial and temporal aggregation, clustering, candidate-source definition, lag, normalisation, and covariate selection" artifact F_sensitivity_results : feature(D_signal_units) value structure: vector value type: {numeric, ordinal} representation form: "cross-specification stability measures" description: "Stability of origin rankings, model support, predictions, residuals, and deployment priority across defensible settings" transform T_assess_hypothesis_evidence : intent: assess manner: "evaluate generalisation, robustness, and mechanism support" input: V_model_diagnostics, P_model_patterns, F_sensitivity_results, S_evaluation, S_sensitivity, S_hypothesis_models, K_transmission_hypotheses, K_roles output: K_model_assessment, S_model_refined actor: human description: "The data analyst and epidemiologist assess held-out performance, residual structure, sensitivity, epidemiological plausibility, and whether a model refinement is justified; they label conclusions robust, sensitive, unsupported, or inconclusive" artifact K_model_assessment : knowledge(F_model_outputs, P_model_patterns, F_sensitivity_results) representation form: "comparative quality judgment" description: "Assessment of model adequacy, generalisation, robustness, residual limitations, and relative support for each mechanism" artifact S_model_refined : specification representation form: "revised model configuration" description: "Revised predictor, lag, interaction, source-definition, or model-complexity settings justified by diagnostics rather than by unrecorded exploratory fitting" if K_model_assessment indicates justified refinement is needed: then: transform T_refine_transmission_models : intent: build-model manner: "residual-driven model revision" input: D_discovery_units, F_outbreak_indicators, F_exposure_predictors, P_model_patterns, F_sensitivity_results, S_model_refined, S_bias_controls, M_transmission_models output: M_transmission_models_refined actor: hybrid description: "Refit the competing models after a documented, evidence-based revision; retain held-out units for re-evaluation" artifact M_transmission_models_refined : model(D_discovery_units, F_outbreak_indicators, F_exposure_predictors) model type: "refined suite of spatiotemporal explanatory models" representation form: "revised mechanism-specific models with explicit controls" description: "Updated competing models justified by residual and sensitivity analysis" assign: M_transmission_models := M_transmission_models_refined S_model_current := S_model_refined else: exit loop L_hypothesis_evaluation end loop L_hypothesis_evaluation

transform T_extract_place_references : intent: characterise manner: "named-entity and location-reference extraction" input: D_clean_posts, F_content_classification, D_city_features, S_governance output: F_place_references actor: hybrid description: "Extract and contextually verify references to landmarks, hospitals, workplaces, events, water locations, and travel locations" artifact F_place_references : feature(D_clean_posts) value structure: list value type: reference representation form: "linked map-feature references and unresolved place names" description: "Place and facility references in microblog content" transform T_visualise_source_investigation : intent: visualise manner: "linked source-evidence investigation" input: D_symptom_reports, D_severe_reports, F_place_references, P_origin_candidates, F_model_outputs, F_sensitivity_results, D_city_features, F_reporting_bias output: V_source_investigation actor: machine description: "Provide temporally ordered source messages, candidate-source maps, matched comparison areas, later-hotspot back-tracing, and conflicting evidence without exposing unnecessary individual-level information" artifact V_source_investigation : visualisation(D_symptom_reports, D_severe_reports, F_place_references, P_origin_candidates, F_model_outputs, F_sensitivity_results, D_city_features) layout: "linked map, aggregated message timeline, candidate-source panel, matched-area comparison, and model-evidence views" form: "aggregated points, text excerpts, ranked lists, contours, and time series" encoding: "Locations and times derive from aggregated reports; overlays show candidate origins, water bodies, hospitals, highways, landmarks, and model support; source text is access-controlled under S_governance" description: "Detailed source-investigation display for epidemiological and environmental field investigation" transform T_formulate_field_validation_plan : intent: generate-knowledge manner: "derive-rules and formulate-statements" input: V_source_investigation, K_origin_assessment, K_model_assessment, K_transmission_hypotheses, K_epidemiology, K_roles output: K_field_validation_plan actor: human description: "The epidemiologist and environmental-health investigator specify targeted clinical verification, facility inspection, water sampling, environmental sampling, and source investigation needed to discriminate among remaining hypotheses" artifact K_field_validation_plan : knowledge(V_source_investigation, K_origin_assessment, K_model_assessment) representation form: "prioritised validation actions" description: "Requests for clinical surveillance, laboratory confirmation, environmental sampling, water-system inspection, and field investigation" if D_confirmation_data contains relevant external evidence: then: transform T_characterise_confirmation_evidence : intent: characterise manner: "align and compare clinical and environmental evidence" input: D_confirmation_data, D_analysis_areas, F_outbreak_indicators, P_origin_candidates, K_field_validation_plan output: F_confirmation_alignment actor: machine description: "Align confirmed cases, clinical activity, laboratory results, and environmental observations with candidate origins, exposure areas, and observed social-media signals" artifact F_confirmation_alignment : feature(D_confirmation_data) value structure: vector value type: {numeric, categorical, temporal} representation form: "agreement, disagreement, and timing-alignment measures" description: "Alignment of external evidence with social-media-derived origin and mechanism hypotheses" transform T_visualise_confirmation_evidence : intent: visualise manner: "comparative clinical-environmental validation display" input: D_confirmation_data, F_confirmation_alignment, P_origin_candidates, F_model_outputs, D_city_features output: V_confirmation_evidence actor: machine description: "Show agreement and contradiction between clinical, environmental, and social-media evidence at appropriate aggregate scale" artifact V_confirmation_evidence : visualisation(D_confirmation_data, F_confirmation_alignment, P_origin_candidates, F_model_outputs, D_city_features) layout: "linked aggregate map, aligned timeline, and hypothesis evidence panel" form: "coloured areas, points, lines, and agreement indicators" encoding: "Spatial and temporal position encode approximate evidence location and observation time; colour distinguishes evidence type, agreement, and contradiction" description: "Validation display for confirming, weakening, or rejecting social-media-derived conclusions" transform T_assess_confirmation : intent: assess manner: "evaluate external confirmation and reanalysis need" input: V_confirmation_evidence, F_confirmation_alignment, K_origin_assessment, K_model_assessment, S_evaluation output: K_confirmation_assessment, S_reanalysis_trigger actor: human description: "The epidemiologist judges whether external evidence corroborates, narrows, contradicts, or is insufficient to evaluate social-media-derived hypotheses; contradictions may trigger a full re-run from signal preparation with updated evidence." artifact K_confirmation_assessment : knowledge(V_confirmation_evidence, F_confirmation_alignment) representation form: "confirmation-status judgment" description: "Assessment of corroboration, contradiction, insufficiency, and evidential limitations" artifact S_reanalysis_trigger : specification representation form: "reanalysis decision rule" description: "Decision on whether contradiction, new evidence, or changed conditions require the workflow to be restarted from signal preparation" else: transform T_record_confirmation_gap : intent: generate-knowledge manner: "formulate-statements" input: K_field_validation_plan, K_model_assessment output: K_confirmation_assessment, S_reanalysis_trigger actor: human description: "Record that no confirming clinical or environmental evidence is available; retain only provisional mechanism statements and specify that later evidence requires reanalysis." artifact K_confirmation_assessment : knowledge(K_field_validation_plan, K_model_assessment) representation form: "confirmation-gap statement" description: "Statement that social-media findings remain unconfirmed and cannot establish a definitive outbreak origin or transmission mechanism" artifact S_reanalysis_trigger : specification representation form: "reanalysis decision rule" description: "Require a full re-run when relevant confirmation data or materially contradictory evidence become available"

transform T_synthesise_transmission_assessment : intent: generate-knowledge manner: "formulate uncertainty-aware explanatory statements" input: K_origin_assessment, K_model_assessment, K_confirmation_assessment, K_transmission_hypotheses, K_signal_preparation_assessment, K_epidemiology output: K_transmission_assessment actor: human description: "The epidemiologist synthesises relative support for person-to-person, airborne, water-related, point-source, reporting-artifact, and mixed explanations. Claims remain explicitly provisional when confirmation evidence is absent or insufficient." artifact K_transmission_assessment : knowledge(K_origin_assessment, K_model_assessment, K_confirmation_assessment) representation form: "ranked uncertainty-aware explanatory assessment" description: "Supported, sensitive, unsupported, and inconclusive transmission explanations with evidence, contradictions, and confirmation status" transform T_record_evidence_ledger : intent: generate-knowledge manner: "write auditable evidence ledger" input: K_signal_preparation_assessment, K_origin_assessment, K_model_assessment, K_confirmation_assessment, K_transmission_assessment, S_hypothesis_models, S_model_current output: K_evidence_ledger actor: human description: "Record conclusions, source artifacts, specifications, contradictory evidence, uncertainty, responsible role, validation status, and outstanding field-investigation needs" artifact K_evidence_ledger : knowledge(K_transmission_assessment, K_origin_assessment, K_model_assessment, K_confirmation_assessment) representation form: "versioned evidence ledger" description: "Auditable record linking operational conclusions to supporting evidence, assumptions, settings, contradictions, confidence, and later validation outcomes" transform T_compute_resource_priority : intent: characterise manner: "equity-constrained risk and deployment prioritisation" input: D_analysis_areas, F_outbreak_indicators, F_reporting_bias, F_model_outputs, F_sensitivity_results, K_transmission_assessment, S_resource_priority, S_governance output: F_resource_priority actor: machine description: "Compute priority using current and growing illness signals, severe reports, population exposure, healthcare context, uncertainty, plausible mechanism support, reporting coverage, and equity constraints" artifact F_resource_priority : feature(D_analysis_areas) value structure: vector value type: {numeric, ordinal} representation form: "priority score and explanatory components" description: "Area-level emergency-response priority with uncertainty and equity-aware explanatory factors" transform T_visualise_operational_dashboard : intent: visualise manner: "linked emergency-management dashboard" input: A_signal_map, F_outbreak_indicators, F_reporting_bias, P_origin_candidates, K_transmission_assessment, F_resource_priority, D_city_features, K_evidence_ledger output: V_operational_dashboard actor: machine description: "Provide linked current-risk, uncertainty, origin-candidate, transmission-evidence, source-investigation, and deployment-priority views" artifact V_operational_dashboard : visualisation(A_signal_map, F_outbreak_indicators, F_reporting_bias, P_origin_candidates, K_transmission_assessment, F_resource_priority, D_city_features) layout: "linked current-risk map, trend views, origin ranking, transmission evidence panel, uncertainty panel, and resource-priority map" form: "coloured areas, ranked lists, time series, symbols, and uncertainty bands" encoding: "Area colour and intensity encode risk and growth; overlays encode uncertainty, reporting coverage, severe reports, hospitals, water bodies, highways, candidate origins, and priority" description: "Operational dashboard for epidemiologists, field investigators, and emergency managers" transform T_formulate_deployment_recommendations : intent: generate-knowledge manner: "derive-rules and write-summary" input: V_operational_dashboard, K_transmission_assessment, K_field_validation_plan, K_evidence_ledger, K_roles, S_resource_priority output: K_deployment_recommendations actor: human description: "The emergency manager, informed by epidemiological and field expertise, recommends immediate but reversible deployment actions, targeted validation, public communication, and safeguards for uncertain or low-coverage areas" artifact K_deployment_recommendations : knowledge(V_operational_dashboard, K_transmission_assessment, K_field_validation_plan) representation form: "prioritised emergency-response recommendations" description: "Recommendations for testing, clinical outreach, field investigation, environmental inspection, communications, and resource allocation with stated uncertainty and equity rationale"

assign: D_surveillance_current := D_signal_units K_deployment_current := K_deployment_recommendations K_ledger_current := K_evidence_ledger loop L_operational_monitoring: purpose: "Monitor subsequent surveillance and field evidence after emergency actions, distinguish likely epidemiological change from reporting change, and update deployment decisions while the incident remains active" until: "The emergency manager and epidemiologist judge that the outbreak response has ended or that ongoing monitoring has been transferred to routine public-health surveillance" body: transform T_integrate_operational_updates : intent: define-unit manner: "append and reconcile updated surveillance observations" input: D_surveillance_current, D_operational_updates, D_confirmation_data, S_governance output: D_surveillance_updated actor: hybrid description: "Integrate newly available aggregate surveillance, confirmation, field-investigation, and intervention information without exposing individual-level records" artifact D_surveillance_updated : entities internal structure: episode embedment: {space, time} description: "Updated area-time surveillance units and associated operational evidence" transform T_characterise_post_intervention_change : intent: characterise manner: "compare trends and reporting conditions before and after actions" input: D_surveillance_updated, F_reporting_bias, K_deployment_current, S_intervention_evaluation output: F_intervention_change actor: machine description: "Estimate changes in symptom activity, severe reports, posting volume, coverage, and confirmation evidence after interventions" artifact F_intervention_change : feature(D_surveillance_updated) value structure: vector value type: {numeric, temporal} representation form: "pre-post trend and reporting-condition comparison" description: "Indicators of epidemiological change, reporting change, intervention timing, and uncertainty" transform T_visualise_operational_change : intent: visualise manner: "pre-post intervention monitoring display" input: D_surveillance_updated, F_intervention_change, K_deployment_current, F_resource_priority output: V_operational_change actor: machine description: "Show intervention timing, area trends, changes in posting and reporting coverage, continuing hotspots, and emerging anomalies" artifact V_operational_change : visualisation(D_surveillance_updated, F_intervention_change, K_deployment_current, F_resource_priority) layout: "aligned before-after timelines and linked priority map" form: "lines, coloured areas, intervention markers, and uncertainty bands" encoding: "Timeline position represents time; markers represent deployed actions; lines show signal and posting-volume changes; map colour shows revised priority or persistent uncertainty" description: "Monitoring view for interpreting response effects cautiously" transform T_assess_operational_change : intent: assess manner: "evaluate intervention effects and reanalysis need" input: V_operational_change, F_intervention_change, S_intervention_evaluation, S_reanalysis_trigger, K_confirmation_assessment, K_roles output: K_operational_assessment, S_operational_reanalysis actor: human description: "The epidemiologist and emergency manager judge whether observed changes likely indicate reduced illness, altered reporting, emerging risk, or insufficient evidence; they determine whether the full analytical workflow must be re-run." artifact K_operational_assessment : knowledge(V_operational_change, F_intervention_change) representation form: "response-effect and adequacy judgment" description: "Assessment of outbreak trajectory, intervention effect, reporting artifacts, emerging needs, and continued uncertainty" artifact S_operational_reanalysis : specification representation form: "full-workflow reanalysis decision" description: "Decision rule specifying whether new evidence, contradiction, or intervention-related changes require restarting the workflow from signal preparation and hypothesis formulation" transform T_update_response_recommendations : intent: generate-knowledge manner: "update recommendations and evidence ledger" input: K_operational_assessment, K_deployment_current, K_ledger_current, S_operational_reanalysis, K_transmission_assessment output: K_deployment_updated, K_evidence_ledger_updated actor: human description: "Update deployment, field-validation, and communication recommendations; append response outcomes, evidence, and reanalysis decisions to the audit ledger" artifact K_deployment_updated : knowledge(K_operational_assessment, K_deployment_current) representation form: "updated prioritised recommendations" description: "Revised emergency-response actions and monitoring priorities" artifact K_evidence_ledger_updated : knowledge(K_ledger_current, K_operational_assessment) representation form: "updated versioned evidence ledger" description: "Updated audit trail including intervention outcomes, new evidence, contradictions, and reanalysis decisions" assign: D_surveillance_current := D_surveillance_updated K_deployment_current := K_deployment_updated K_ledger_current := K_evidence_ledger_updated end loop L_operational_monitoring