{"id":"distributed-specialization","title":"Distributed specialization and integration","regions":["distributed-learning","collective-structure"],"question":"How can smaller specialized learners contribute to a larger learning system without incompatible updates accumulating?","formulation":"The proposed arrangement distributes learning across heterogeneous resources, while management and integration can themselves be assigned to models. Coarse and fine representations offer one possible division of labor.","representations":[{"name":"Specialists contributing knowledge artifacts","affords":"Lets contributions be inspected before integration.","loses":"Requires assessment and reconciliation.","epistemic_kind":"reconstruction"},{"name":"A manager distributing adaptive learners","affords":"Coordinates specialization and resource allocation.","loses":"Concentrates decisions and can create a bottleneck or shared blind spot.","epistemic_kind":"reconstruction"}],"tension":"Distribution of computation does not necessarily distribute interpretive or political control.","challenge":"Who decides whether a local improvement is a valid contribution to the shared model?","development":"active","epistemic_kind":"reconstruction","evidence":["basis-distributed-specialization"],"next_move":{"text":"Run a small conflicting-update example and compare integration rules using both local and shared task performance.","epistemic_kind":"model_proposed"},"open_questions":["How can smaller specialized learners contribute to a larger learning system without incompatible updates accumulating?","Who decides whether a local improvement is a valid contribution to the shared model?"],"formal_status":"open_problem","version":1,"content_hash":"ca5ad53d9e84667e0f7a4f3e5210c3653a80ce24e74c1bac9c8eaa3c552fe8e8"}