A Review of Cognitive Decision-making within Future Mission Systems

This paper provides an outline of historical attempts to achieve human-like decision-making within machines. It concludes with a proposed conceptual approach of how researchers might pursue cognitive mission systems designs in the future. A number of potential success stories need to be explored in order to revise existing techniques and identify which techniques could be componentised for use in this future design. Existing cognitive systems have evolved over time, using; LISt Pro- cessing (LISP), PROLOG and Object Oriented Programming (OOP) languages. These were used to represent information using lists, scripts, frames, schemas, production rules, procedural, semantic and declarative processes. As Computational Intelligence (CI) techniques evolved, a number of frameworks emerged; such as Recognition-Primed Decision (RPD), Pro- cedural Reasoning System (PRS), Collaborative Agent for Simulating Teamwork (CAST), Adaptive Character of Thought– Rational (ACT-R), distributed Multi-Agent Reasoning System (dMars), State, Operator And Result (SOAR) and Java Agent Compiler and Kernel (JACK). Many of these failed to gain traction because the problem-space has become more complex and existing heuristic code quickly becomes unwieldy with no guaranteed solution. Although currently heuristic systems relieve humans of routine activities, they are not able to independently reproduce intuition, insight or cognitive learning. Researchers have repetitively attempted to enhance the level of decision-making capabilities, but few have achieved success without aug- mented human support. Emerging frameworks continue to re-use a number of recurring themes to solve constrained problems, although most techniques cannot transform information into knowledge or wisdom. This paper highlights a number of the more successful concepts that could be used to progressively derive components to form a working cognitive decision-making model within a future mission system