AI-enabled, user-modifiable, systems simulation web apps

Scientists who use simulation models in their research, or teach dynamical systems concepts, are faced with the challenge of what modeling platforms to use. Platforms, such as Mathematica, Matlab, Anylogic, or R require dedicated training courses to become proficient in their use, while other lighter weight platforms, such as Stella or Netlogo also require concentrated training to effectively use them. Here we present AI-enabled, user-modifiable, simulation Web apps that can be accessed directly online for immediate and free use. These Apps have been constructed using NuWebKit TM technology, which allows models to be run directly in a browser or on a remote server. Additionally, NuWebKit TM models can be integrated as procedures/packages to be used on other platforms (e.g., R Studio). First, we present details of the NuWebKit TM architecture that enables these features. The flexibility of NuWebKit TM Apps derives from Numerus’ RAM (Runtime Alterable Module; not to be confused with memory RAM) technology, which allows users to modify or insert equations on-the-fly, using an intuitive, easy-to-master syntax. It also includes Apps with an AI-enabled window in which the user can instruct AI to build a particular simulation model and provide an accompanying user-document that appears in a second window on the Apps user-interface. Second, we discuss specific NuWebKit TM Apps, available at a Numerus Webkit Examples webpage, for simulating population epidemics, animal movement, and ecological processes. Finally, for the first time we present details of the NuWebKit TM DynSS App. This is a general dynamical systems app for simulating systems of differential and discrete-time deterministic or stochastic models whose equations are either inserted by the user into the app’s equation coding window, or generated through an embedded AI facility. In both user-coding or AI-mode the DynSS App automatically populates a run-time dashboard of sliders for selecting model parameter values, graphs for visualizing solutions, and facilities to carry out sensitivity studies and stochastic simulations. These features are illustrated in the context of the Hodgkin-Huxley neuron activation model, a discrete-time, age-structured, stochastic harvesting model, and a population viability analysis model.