A Methodology and Software Environment for Testing Process Model's Sequential Predictions with Protocols
Abstract
Getting the most out of information processing models requires that testing and refining them be straightforward. This requires that (a) large amounts of data be easily compared with the model's performance, (b) descriptions of how and where the model mismatches are readily available and easy to interpret, and (c) the models themselves can be refined in a straightforward way. Current methods for testing the sequential predictions of process models provides none of these. It is a difficult, time consuming, boring task, requiring the full attention of a skilled analyst Despite the importance and and difficulty of testing process models against protocol data, and in contrast to the rich methodology for analyzing samples of numerical data, there is no explicit methodology or set of tools for automatically or semi- automatically doing this task. This thesis specifies a methodology for testing process models sequential predictions through comparison with verbal and non- verbal protocols. Each of its steps are delineated, and the requirements to perform these steps developed.... Spreadsheets, Programmer workbench, Program editors, Tracing, Display algorithms, Help and documentation, Simulation support systems, Expert system tools and techniques, Model development, Model validation and analysis, Cognitive simulation, Relations among models, Protocol analysis, Soar.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 21, 1992
- Accession Number
- ADA260980
Entities
People
- Frank Ritter
Organizations
- Carnegie Mellon University