WHERE? Risk, Uncertainty and Software Safety

J A McDermid

WHERE? Risk, Uncertainty and Software Safety

Computer Science

Research output: Contribution to conference › Paper

Abstract

It is widely accepted that safety is concerned with risk, defined as a product of probability and severity; however we tend not to probe too deeply what is meant by probability, nor to investigate its roots in uncertainty. Broadly we can evaluate risk in the classical manner when we understand the probabilities or likelihoods of possible events. We have uncertainty, however, when we know about possibilities but not probabilities. Generally, when dealing with software, it is practicable to identify possibilities, e.g. undesired behavior, but not probabilities of occurrence, hence the difficulties which have arisen in trying to integrate software into classical safety processes. The outlines the concepts of probability and uncertainty, and propose a means of producing systematic arguments about software safety – by considering the role of evidence in reducing uncertainty. The approach draws on some of Keynes’ early work in economics. The proposal leads naturally to an evidence-based approach to software safety cases, where some of the arguments reflect the weight or confidence that can be placed on some item of evidence in showing the absence of some undesired behavior.

Original language	English
Publication status	Published - 2008

Cite this

@conference{2cbae6bd15a34a27833ab40a11fd0dd4,

title = "WHERE? Risk, Uncertainty and Software Safety",

abstract = "It is widely accepted that safety is concerned with risk, defined as a product of probability and severity; however we tend not to probe too deeply what is meant by probability, nor to investigate its roots in uncertainty. Broadly we can evaluate risk in the classical manner when we understand the probabilities or likelihoods of possible events. We have uncertainty, however, when we know about possibilities but not probabilities. Generally, when dealing with software, it is practicable to identify possibilities, e.g. undesired behavior, but not probabilities of occurrence, hence the difficulties which have arisen in trying to integrate software into classical safety processes. The outlines the concepts of probability and uncertainty, and propose a means of producing systematic arguments about software safety – by considering the role of evidence in reducing uncertainty. The approach draws on some of Keynes{\textquoteright} early work in economics. The proposal leads naturally to an evidence-based approach to software safety cases, where some of the arguments reflect the weight or confidence that can be placed on some item of evidence in showing the absence of some undesired behavior.",

author = "McDermid, {J A}",

year = "2008",

language = "English",

}

TY - CONF

T1 - WHERE? Risk, Uncertainty and Software Safety

AU - McDermid, J A

PY - 2008

Y1 - 2008

N2 - It is widely accepted that safety is concerned with risk, defined as a product of probability and severity; however we tend not to probe too deeply what is meant by probability, nor to investigate its roots in uncertainty. Broadly we can evaluate risk in the classical manner when we understand the probabilities or likelihoods of possible events. We have uncertainty, however, when we know about possibilities but not probabilities. Generally, when dealing with software, it is practicable to identify possibilities, e.g. undesired behavior, but not probabilities of occurrence, hence the difficulties which have arisen in trying to integrate software into classical safety processes. The outlines the concepts of probability and uncertainty, and propose a means of producing systematic arguments about software safety – by considering the role of evidence in reducing uncertainty. The approach draws on some of Keynes’ early work in economics. The proposal leads naturally to an evidence-based approach to software safety cases, where some of the arguments reflect the weight or confidence that can be placed on some item of evidence in showing the absence of some undesired behavior.

AB - It is widely accepted that safety is concerned with risk, defined as a product of probability and severity; however we tend not to probe too deeply what is meant by probability, nor to investigate its roots in uncertainty. Broadly we can evaluate risk in the classical manner when we understand the probabilities or likelihoods of possible events. We have uncertainty, however, when we know about possibilities but not probabilities. Generally, when dealing with software, it is practicable to identify possibilities, e.g. undesired behavior, but not probabilities of occurrence, hence the difficulties which have arisen in trying to integrate software into classical safety processes. The outlines the concepts of probability and uncertainty, and propose a means of producing systematic arguments about software safety – by considering the role of evidence in reducing uncertainty. The approach draws on some of Keynes’ early work in economics. The proposal leads naturally to an evidence-based approach to software safety cases, where some of the arguments reflect the weight or confidence that can be placed on some item of evidence in showing the absence of some undesired behavior.

M3 - Paper

ER -