{
  "field_id": "F013",
  "field": "affective computing",
  "directions": [
    {
      "direction_id": "G0",
      "name": "Visual inference of depicted people's affect",
      "scope": "Infer individual or group emotions from depicted people's faces, actions, and scene and social context, including joint learning of related facial affect descriptors."
    },
    {
      "direction_id": "G1",
      "name": "Affect inference from communicative signals",
      "scope": "Infer expressed emotion or affect from speech, non-linguistic vocalizations, text, and combined audio, visual, and textual evidence, including the use of conversational relationships and turn context."
    },
    {
      "direction_id": "G2",
      "name": "Affect elicited by media",
      "scope": "Predict or explain the emotional responses that images, videos, memes, art, music, and other content elicit in viewers or listeners, including response distributions and reasons for those responses."
    },
    {
      "direction_id": "G3",
      "name": "Monitoring affective and cognitive states",
      "scope": "Estimate people's emotions and activity-related stress, engagement, attention, fatigue, and cognitive load from physiological, neural, thermal, or behavioral observations."
    },
    {
      "direction_id": "G4",
      "name": "Learning with limited or uncertain affect supervision",
      "scope": "Obtain representations and supervision usable for affect prediction when affect annotations are scarce, ambiguous, or unreliable, and address label disagreement or predictive uncertainty."
    },
    {
      "direction_id": "G5",
      "name": "Affect inference across shifts and impaired observations",
      "scope": "Maintain effective affect inference when populations or corpora change, or when expected affective evidence is missing, occluded, noisy, or degraded."
    },
    {
      "direction_id": "G6",
      "name": "Fairness of affective inference",
      "scope": "Identify or mitigate demographic or other group-dependent bias and unfairness in the inference of affect and related psychological states."
    },
    {
      "direction_id": "G7",
      "name": "Explaining emotions through events and roles",
      "scope": "Link emotions to the events that elicit them, relevant participants, or cognitive appraisals of situations to extract or explain the basis on which emotions arise."
    },
    {
      "direction_id": "G8",
      "name": "Emotionally responsive agents",
      "scope": "Enable interactive agents to reason, provide support or interventions, or respond with embodied affect in light of people's emotions or circumstances."
    },
    {
      "direction_id": "G9",
      "name": "Emotion-controlled expressive generation",
      "scope": "Generate or transform music, speech, images, and facial reactions so that the outputs convey an intended emotion or have an appropriate affective expression."
    },
    {
      "direction_id": "G10",
      "name": "Affect research resources and measurement validity",
      "scope": "Establish annotated data, tasks, and benchmarks for affect research, and examine whether labels, measurements, and evaluation protocols support credible research conclusions."
    }
  ]
}