{
  "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."
    }
  ]
}