SOFTWARE AND HUMAN RELIABILITY: ERROR REDUCTION AND PREDICTION

摘要

The reliability of software is improved when humans detect and correct errors in the coding of logic, inputs, and programming during the various stages of the engineering lifecycle, including development, testing and operation. We postulate that these methods are the name as, and reflect the neural learning processes that dominate theories of human reliability. Our previous research has demonstrated thai new dynamic laws derived from the learning hypothesis govern the observed behavior of humans in the process of learning and correcting errors. The data used to validate this theory were classic data sets taken from the cognitive psychology literature for error correction and response times. Developing the mental skills of error correction, learning, decision making, and problem solving all reflect the neural connectivity and patterns that correspond to the emergence of order from disorder and randomness. The statistical methods and probabilistic distributions that can characterize these behaviors arise naturally from the processing of complexity as measured by the information entropy. We examine software reliability and testing data, using learning theory to demonstrate with both data and theoretical analysis that there is indeed a direct and proportional relation between response times for processing complexity and correcting mistakes by developing learning patterns. Our results indicate that response times decrease slightly faster than error correction rates. We also demonstrate a high degree of consistency between individual learning, error correction, skill acquisition rates, and responses with strong trends exhibited through learning and error reduction during the testing of large software applications as well as their operation in field environments.