This textual analysis examines computer simulations as rhetorical objects and acts. In particular, this work examines scientific simulations from organic chemistry and astrophysics in order to expose how rhetorical and social aspects influence the ad hoc decisions (e.g., setting initial parameters, excluding and adding arbitrary elements, and making other choices) that comprise simulations. Prior works in philosophy, critical theory and technical communication underscore fictional and formal features of simulation. In contrast, this dissertation dissects multiple levels of documents surrounding actual simulations—not only drafts of published articles but also software and code interiors, e-mail and letter correspondence, newsletters and white paper reports—in order to discuss the relational (rather than purely formal) meaning of the simulations. This work also compares simulation to other modes of the scientific imagination—paradox, thought experiments and metaphor, in particular. My findings suggest that simulations hinge upon abductive (rather than deductive or inductive) reasoning and qualify as virtual evidence. Also, while published drafts of simulation articles tidy the ad hoc twists and turns necessary for creating simulations, prior drafts and peripheral documents attest to the fact that organizational affiliations, earlier projects, and rhetorical strategies help establish the scope and meaning of simulation projects. Further, meaning-making takes place well before and long after the article drafting process—in prior incarnations of the work for presentation, in correspondence between article writers and reviewers, and in citations in others’ writing. / text

As scientific simulations get increasingly complex, so do the requirements of how to deal with the data that is produced. Few scientists and engineers today are satisfied with just looking at streams of numbers; we require graphical visualizations to better understand their meaning. The traditional method of visualization has been to save the simulation's results to a file, then load that file up in another program (eg. Microsoft Excel) for post-processing. Although post-processing data to produce visualizations may be sufficient for some simple simulations, a modern simulation designer usually wants more out of their visualization. Perhaps they want the visualization to be a 3D plot of an interplanetary trajectory, with the ability to zoom, pan, and rotate the scene interactively. Until now, doing so has required the designer to become adept at computer graphics, which is a feat that almost no scientist or engineer has the time to attempt. The research undertaken here introduces an architecture by which a simulation programmer can easily add interactive 3D visualizations to their simulations. This architecture has several benefits over existing visualization packages, the biggest one being that no knowledge of computer graphics is required to use the it in one's own simulations. Another benefit is that the resulting visualization is interactive by default, without any extra programming required on the part of the simulation designer. This thesis begins by introducing the theory behind how scientific simulations want to visualize data. Common aspects of all simulations are identified, and are used to develop a common "visualization language" that can be used by any simulation designer to specify what they want to visualize. The second part of the thesis specifies a particular implementation of this visualization language, called OpenFrames. Open- Frames is a library of functions that can be called from C, C++, or FORTRAN, and automatically implements the visualization specified by the designer.