Autonomous FAIR Digital Objects: From Passive Assertions to Active Knowledge
Researchers have advanced the concept of Autonomous FAIR Digital Objects (aFDOs) from an abstract idea to an operational model, aiming to transform passive scientific publications into active, self-managing knowledge units. Current web-based scientific data relies on centralized middleware and institutional continuity, often becoming stagnant when registries close. The proposed aFDO framework augments standard FAIR Digital Objects with three capabilities anchored in Semantic Web standards: a policy layer for portable condition-action rules, an announcement layer to bound evaluation costs, and an agreement layer to resolve contradictions using reputation-weighted consensus. The team evaluated an open reference implementation using 4,305 FDOs based on rare-disease ontologies like ClinVar, HPO, and Orphanet. Results showed the consensus mechanism successfully resolved 56.3% of naturally occurring conflicts in ClinVar data where expert panels had previously adjudicated. Furthermore, the system demonstrated graceful degradation under Sybil, collusion, and poisoning attacks within its designed Byzantine-tolerance bounds. This development offers a pathway toward accountable, standards-aligned automation in scientific data stewardship that can outlive the institutions that originally published the data.
Wire timeline
Autonomous FAIR Digital Objects: From Passive Assertions to Active Knowledge
Researchers have advanced the concept of Autonomous FAIR Digital Objects (aFDOs) from an abstract idea to an operational model, aiming to transform passive scientific publications into active, self-managing knowledge units. Current web-based scientific data relies on centralized middleware and institutional continuity, often becoming stagnant when registries close. The proposed aFDO framework augments standard FAIR Digital Objects with three capabilities anchored in Semantic Web standards: a policy layer for portable condition-action rules, an announcement layer to bound evaluation costs, and an agreement layer to resolve contradictions using reputation-weighted consensus. The team evaluated an open reference implementation using 4,305 FDOs based on rare-disease ontologies like ClinVar, HPO, and Orphanet. Results showed the consensus mechanism successfully resolved 56.3% of naturally occurring conflicts in ClinVar data where expert panels had previously adjudicated. Furthermore, the system demonstrated graceful degradation under Sybil, collusion, and poisoning attacks within its designed Byzantine-tolerance bounds. This development offers a pathway toward accountable, standards-aligned automation in scientific data stewardship that can outlive the institutions that originally published the data.
cs.AI updates on arXiv.org