| { |
| "field_id": "F030", |
| "field": "causal machine learning", |
| "directions": [ |
| { |
| "direction_id": "G0", |
| "name": "Latent causal modeling and counterfactual generation", |
| "scope": "Establish causally meaningful latent variables, mechanisms, or observation mappings from complex observations, and study the interventions and counterfactual observations supported by these models." |
| }, |
| { |
| "direction_id": "G1", |
| "name": "Causal structure discovery", |
| "scope": "Infer or disambiguate causal directions, relations, and structures among variables using evidence from observational statistics, temporal dynamics, interventional environments, or knowledge provided by language models." |
| }, |
| { |
| "direction_id": "G2", |
| "name": "Causal-effect identification and bounds", |
| "scope": "Determine the extent to which available distributions and causal assumptions establish a target causal quantity, and provide an identifying expression or a range when point identification is not possible." |
| }, |
| { |
| "direction_id": "G3", |
| "name": "Causal-effect estimation across data and treatment settings", |
| "scope": "Use nonrandomized observations, auxiliary variables for identification, longitudinal records, donor or external-population evidence, and data from interacting units to estimate treatment outcomes, effects, and counterfactuals for target populations or units." |
| }, |
| { |
| "direction_id": "G4", |
| "name": "Causal intervention and sequential policy learning", |
| "scope": "Use information about causal effects or mechanisms to choose actions and learn allocation rules, sequential interventions, or agent policies in environments that improve outcomes." |
| }, |
| { |
| "direction_id": "G5", |
| "name": "Causal experiment design and inference", |
| "scope": "Design experiments or analyze experimental data under randomized assignment, adaptive implementation, and sampling constraints to obtain valid and precise causal estimates." |
| }, |
| { |
| "direction_id": "G6", |
| "name": "Validation of causal models, methods and reasoning", |
| "scope": "Examine the reliability of causal models, estimation methods, or causal judgments made by language models, and analyze the conditions under which they apply or fail." |
| }, |
| { |
| "direction_id": "G7", |
| "name": "Causally robust prediction", |
| "scope": "Use causal mechanisms, invariance, or interventional information to keep predictors effective under distribution shifts and spurious associations." |
| }, |
| { |
| "direction_id": "G8", |
| "name": "Causal fairness", |
| "scope": "Define and improve the fairness of predictions and action allocation using causal disparities, counterfactuals, or intervention benefits." |
| }, |
| { |
| "direction_id": "G9", |
| "name": "Causal explanation and root-cause diagnosis", |
| "scope": "For a specific output or an anomaly that has occurred, determine which factors or changes caused the result and provide interpretable causal attribution." |
| }, |
| { |
| "direction_id": "G10", |
| "name": "Intervention-preserving causal abstraction", |
| "scope": "Simplify causal models or relate different levels of abstraction while preserving the required consistency of interventional behavior." |
| } |
| ] |
| } |
|
|
