Knowledge workers in the DOD and IC enterprise require human-centered workflows that enable human-IT partnerships to ensure data quality and utility in the operational environment. To turn the ISR data into actionable intelligence at operational speed, analysts look to utilize the automated deep learning (DL) algorithms which showed success on a variety of pattern-recognition tasks. However, current DL applications fail to explain why the system made choices that it did to help common users make sense of DL behavior. This detracts from implementing synergistic interactive human-computer collaboration and precludes effective curation by SME of DL algorithms.To address the problems of devising human-centric workflows for easy-to-operate expert-guided machine-learning and augmenting human capacities with trustworthy, cognitively-friendly intelligent automation aids for synergistic human-automation collaboration to rapidly extract and generate high-value intelligence, we are developing SCHEMA (Synergistic Curation by Humans Experts of intelligible Machine-learning Apps). SCHEMA implements and integrates the semi-automated human-centered TCPED processes/workflows for the three enabler dimensions of human-automation interaction: (1) Cognitively-friendly machine-to-human input with self-explanatory, self-reconfiguring DL aids; (2) Man- and machine-readable unifying knowledge graphs for man-machine dialogue and low-workload human supervision, feedback, and curation of DL models; and (3) Synergistic human-automation collaboration.Cognitively-friendly automation,Knowledge graphs,Man-machine dialogue,Workflows,Human-IT collaboration,Feedback,Training,Deep Learning models
