{
  "field_id": "F017",
  "field": "algorithmic fairness",
  "directions": [
    {
      "direction_id": "G0",
      "name": "Fairness requirements and measurement validity",
      "scope": "Determine which fairness goals are justified in context, and examine the validity of the measures, group attributes, and interpretations used to express disparities or social bias."
    },
    {
      "direction_id": "G1",
      "name": "Detection, diagnosis and assurance of unfair behavior",
      "scope": "Determine whether a system exhibits unfair behavior, who is affected, and why disparities arise, providing verifiable fairness conclusions or ongoing monitoring results when needed."
    },
    {
      "direction_id": "G2",
      "name": "Representational fairness of generated content",
      "scope": "Identify demographic stereotypes and bias in generated language, images, or multimodal content, and change the generation process or its outputs to reduce them."
    },
    {
      "direction_id": "G3",
      "name": "Fair predictive learning and deployment",
      "scope": "Develop, adjust, or test models and their learning foundations for fair prediction, so that required group, individual, causal, or calibration properties are satisfied or maintained under the relevant data conditions."
    },
    {
      "direction_id": "G4",
      "name": "Fair distribution of positions, membership and actions",
      "scope": "Fairly allocate valuable positions, eligibility for selection or grouping, and actions and limited resources in sequential decisions, while accounting for utility or relevance."
    },
    {
      "direction_id": "G5",
      "name": "Fairness under decision-induced feedback",
      "scope": "Study or control long-term group disparities that arise when deployed decisions change future outcomes, observable labels, or participation."
    },
    {
      "direction_id": "G6",
      "name": "Privacy-compatible fairness",
      "scope": "Characterize and address the constraints between privacy guarantees, risks of disclosing sensitive attributes, and fairness goals when learning or auditing for fairness."
    },
    {
      "direction_id": "G7",
      "name": "Evaluation and operationalization of fairness interventions",
      "scope": "Develop evidence, evaluation procedures, and organizational support for choosing and using fairness interventions in practice, enabling fairness work to move from comparing methods to sustained practice."
    }
  ]
}