Presentation
Toward a Human-Agent-Environment Ontological Framework for AI-Enabled Mission Readiness
SessionThursday Poster Session
DescriptionMission readiness in modern operational settings requires dynamic cooperation between human operators and autonomous agents under uncertainty. In such contexts, human and AI agents are expected to independently analyze information from, select actions within, or directly modify operational environments such as cyber infrastructures, databases, and coding platforms. Existing human factors research and AI governance frameworks typically examine these components in isolation, providing limited support for understanding how mission degradation or failure emerges from their interactions over time. Thus, this paper presents preliminary results from a DARPA-funded effort to develop a three-dimensional ontology for AI-enabled operations. We organize readiness-relevant constructs across Human, Agent, and Environment classes developed through synthesis of empirical and theoretical literature, AI evaluation frameworks, governance standards, and participatory workshops with operational stakeholders. At present, the ontology contains over 200 structured entities spanning agent autonomy and transparency, human workload and trust calibration, and environmental properties such as state persistence, recoverability, and consequence severity. Across these individual dimensions, we explicitly model cross-entity relations (e.g., amplification, constraint, and compensation) to guide the design and evaluation of human-AI interactions. We briefly discuss ongoing instantiations of the ontology within DoD-relevant application domains, including intelligence analysis, cyber-social operations, and multi-UAV reconnaissance missions.
Event Type
Poster
TimeThursday, October 22nd5:30pm - 6:30pm PDT
Location