STELLA (programming language)

While working at the Massachusetts Institute of Technology in the 1960s, Jay Wright Forrester developed the earliest understanding of system dynamics which he argued could only be understood using models. Dartmouth College systems science professor Barry Richmond founded High Performance Systems in 1984. With the financial support of Analog Devices, Inc. and technical support from Apple Computer, he developed STELLA (short for Structural Thinking, Experimental Learning Laboratory with Animation) at his company. He presented the prototype for the visual programming language in 1985 at the System Dynamics Society's annual conference in a paper entitled "STELLA: Software for Bringing System Dynamics to the Other 98%". Steve Peterson, a colleague of Richmond's, reflected after his death in 2002 that Richmond held the belief that modeling was a tool everyone should be using and that that notion was reflected in Richmond's work. He quoted a 1994 paper in which Richmond described STELLA as "quite unique, quite powerful, and quite broadly useful as a way of thinking and or learning. It's also capable of being quite transparent–leveraging the way we learn biology, manage our businesses, or run our personal lives". ==Functionality and features==

STELLA's approach to modeling systems shares some similarities with a precursor, the DYNAMO simulation language. DYNAMO explicitly defined "stocks" (reservoirs) and "flows" (inputs and outputs) as key variables in a system, a vocabulary that STELLA shares. Within STELLA, users are presented with a graphical user interface in which they may create graphical models of a system using four fundamentals: stocks, flows, converters, and connectors. Relationships between converters (which convey transforming variables) and other elements may be drawn with converters. Users are able to input values for stocks, flows, and converters (including a variety of built-in functions). STELLA does not differentiate between external and intermediate variables within a system; all of them are represented with converters. The software produces finite difference equations that describe the graphical model and allows users to select a numerical analysis method to apply to the system, either the Euler method or various Runge–Kutta methods (either second or fourth order). Before running a model, users may also specify a time step and runtime for the simulation. STELLA can output data in graphical or tabular forms. STELLA runs one window at a time, meaning that only one model can be run at any given moment. In 2012, two researchers released StellaR, software which can translate STELLA models into the R programming language. ==Applications==

Education Because of its simplicity relative to more complex modeling languages, STELLA has been cited as a useful tool in educational settings. and modeling the protagonist's motivation throughout William Shakespeare's Hamlet. A 2010 study of the efficacy of project-based learning upon a watershed-modeling project undertaken by 72 middle schoolers found that the addition of a STELLA modeling component in the project improved overall comprehension of the material over traditional methods, especially among female students who outperformed their male counterparts with the addition of STELLA. Academia and commerce The software is also used in research settings. Among other projects, researchers have used STELLA to apply Hubbert peak theory to the Chinese coal supply, to model atrazine dynamics within agricultural lands, and to simulate the interactions between marine macroinvertebrates. isee systems markets an identical software targeted at business consumers under the name iThink (previously STELLA for Business). coordination between an emergency room and hospital beds, and competition in the home video market. ==Reception==

In a 1987 review of the program in BioScience, Robert Costanza wrote that "STELLA is a solid program–well planned and executed–that breaks new ground." A 1991 review of iThink in Planning Review noted that the software's strength was in its low cost (the program retailed at around $450 (US$ in dollars ) and the vast number of possibilities allowed by its open-ended form. ==See also==