As mentioned by the retired U.S. intelligence officer David Shedd, former Deputy Director of the Defense Intelligence Agency (DIA): ”Every successful military plan and operation relies on intelligence.”1 The vast, and quickly growing, amount of data available to analysts from all intelligence sources as well as the quicker pace that intelligence products are required today has created numerous data processing and analysis (as well as classification and compartmentalization of intelligence products amongst Nations) challenges. Given technology developments in all ranges of autonomous systems:
• Human-In-The-Loop (HIL) or Semi-autonomous Systems (Under control or supervision of human operator and are not allowed to implement their critical algorithms outputs (decisions) without approval),
• Human-on-the-loop (HOL) or Supervised-Autonomous Systems –Allowed to implement their decisions; but the human operator has override rights to stop or change performed actions in case of error/failure,
• Human-out-of-the-loop (HOOL) or Fully Autonomous Systems – Implement their decisions without human operator approval; act according to their programming, but are not restricted by any external intervention,
it is important to explore their potential in all areas. As autonomy “at rest” technologies become more mature, it is important to explore the potential of this technology in the context of intelligence.
Analysis of the potential of the different autonomy levels in the four phases of the intelligence cycle (Direction, Collection, Processing and Dissemination). In particular which level of autonomy is required and desirable in a given context, for instance:
o Direction: can autonomy help to develop the necessary intelligence requirements given the mission goals?
o Collection and Processing:
• Fusion of data from multiple sources that takes into account “data triangularization” (cross-checking of all sources of data) to reduce deception
• Processing of data that include Anomaly detection/pattern recognition as well as classification of intel products
o Dissemination: Can autonomy help with interoperability of data sharing and fusion within NATO – security is the key issue in this process
• Mapping of the nature of the teaming relationship required to be established in the intelligence processes – where is the human in the process and where is the machine in the process?
• Development and execution (with operators) of a CDAG to develop new intelligence autonomy concepts and to explore and analyse the potential of autonomy in its different levels in the intelligence cycle.
• The activity will focus on the intelligence process and implementation methodologies so as to avoid getting into classified areas. The activity will remain at the unclassified level.