Simulation analytics as an eco-systemic approach to the nonlinear dynamic response of artificial adaptive systems: A toolkit

This paper introduces Simulation Analytics (SA), an approach that enables comprehensive data exploration through simulation. SA can extract representative profiles from existing data patterns (like typical gang member characteristics), analyze counterfactual scenarios (such as profiles that bridge opposing groups), and even examine logically contradictory cases (like simultaneous membership in mutually exclusive categories). We present several algorithms and architectures that form a basic SA toolkit, demonstrating their application on the widely studied Gangs dataset. Through this analysis, we show how SA offers an eco-systemic approach to machine learning that contrasts with conventional deep learning methods. Rather than relying on single, complex neural networks, SA employs a community of different machine learning components working together. This design makes SA both more versatile in handling various simulation tasks and potentially more representative of biological cognition processes.