A framework for truth and honesty in large language processing for safety engineering

Research Unit for Maritime Health and Technology (MHT), University of Southern Denmark

This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

The work presented here had its origin in the 1970’s [7] when it was shown that a significant fraction of incidents and minor accidents in nuclear plant arose due to errors in operating procedures. Methods were developed for translating operating procedures to a form which supported simulation, allowing detection of errors and omissions automatically [7]. Since then, the methods have been used for quality assurance of risk analyses and safety engineering for over 100 large process plants [9, 13]. This paper describes the development of a framework for a use of grammatical and semantic rule-based methods to support large language models is a way that allows true and honest answers to questions and requests posed, and its implementation in a program, Cassandra, developed by the author.