麻豆传媒

TRAGEDIES like the 1987 fire at King鈥檚 Cross station on the London underground could be prevented if engineers thought more carefully about the language they used in safety documents, says a computing specialist. And the way to do this, he says, might be to borrow one of the techniques of literary criticism.

At the inquiry into the King鈥檚 Cross disaster, which killed 31 people, managers were criticised for using the word 鈥渟moulderings鈥� to refer to small, rubbish-fuelled fires like the one that led to the disaster. This is no isolated case, says Jim Armstrong, a computer scientist specialising in software reliability at the University of Newcastle upon Tyne: 鈥淚t鈥檚 a good example of a euphemism used to pass safety problems off as perfectly natural and ordinary 鈥� not something to worry about.鈥�

Armstrong thinks that the theories of Jacques Derrida, the French philosopher behind the branch of literary criticism called deconstruction, could help uncover ambiguity, bias and self-deception in the safety systems of installations such as nuclear power stations and chemical plants.

A major aim of the project is to find ambiguous terminology that regularly crops up and bring it out into the open. But the very nature of deconstruction means that a formal system for applying it to safety is unlikely to be developed. 鈥淚t cannot settle an argument, only 鈥榰nsettle鈥� it,鈥� Armstrong says.

However, he hopes his work will lead to a new discipline of 鈥渟afety philosophy鈥�, in which the way people think about safety is properly discussed. He thinks the 鈥渞elentlessly sceptical鈥� approach characteristic of deconstruction could help combat the kind of complacency that led to the King鈥檚 Cross fire, or the Chernobyl disaster.

Though Armstrong鈥檚 ideas have, unsurprisingly, had a mixed reception, some safety experts think the approach could be useful. 鈥淚t鈥檚 an unusual, even oddball approach, but it鈥檚 certainly worth looking into,鈥� says Martyn Thomas, an expert in safety-critical software and chairman of a dependable systems research group in Britain.

Definitions of deconstruction theory often run to several pages, but one of the central ideas is that meaning should never go unquestioned. The author鈥檚 words are pulled apart to reveal all other possible meanings, to see if there are any unconscious assumptions at work that may skew things dangerously. 鈥淗e really looks into the limits that language imposes on thought,鈥� Armstrong says.

One concept that is viewed pretty much as holy writ among safety engineers, for example, is that the risks in any system should be reduced to a level that is 鈥渁s low as reasonably practicable鈥�, known as ALARP for short. But, applying Derrida鈥檚 ideas, 鈥渞easonably鈥� simply means 鈥渋n accordance with someone else鈥檚 preconceptions鈥�, says Armstrong.

And 鈥減racticable鈥� boils down to what you鈥檙e willing to pay for. It hides the fact that spending more money, for instance, might mean a safer system. 鈥淥ne of the Derridian insights I鈥檝e had in this project is that to modify a word by putting an 鈥榓ble鈥� on the end of it makes it look specialist and technological, providing an unwarranted degree of comfort,鈥� he says.

Armstrong is in the middle of a two-year project, funded by Britain鈥檚 Engineering and Physical Sciences Research Council, to demonstrate that this sort of approach could help improve safety. He is looking at case studies where things have gone wrong, and hopes to show that deconstruction analyses could have helped prevent them. For legal reasons, he was reluctant to discuss the details of specific cases with 麻豆传媒.

So what has language got to do with the safety of nuclear plants? Anyone designing a safety-critical system to control such a plant has to supply a written 鈥渟afety case鈥� explaining its operation to the relevant regulator, such as Britain鈥檚 Nuclear Installations Inspectorate . This document details circumstances in which the system will operate safely, when it will fail safely, and under what conditions it might risk catastrophe. 鈥淪afety cases are hard to prove correct,鈥� Thomas says.

And according to Armstrong, the so-called 鈥渄ependability arguments鈥� that make up a safety case are utterly subjective 鈥� informed by politics, economics, the writer鈥檚 expertise and potentially limited data on the system itself 鈥� and can contain dangerous linguistic anomalies. For instance, if an engineer states that a nuclear waste tank is safe as long as the level of the noxious liquor it contains never exceeds a certain level for more than an hour, what do they actually mean by 鈥渢he level鈥�? Do they mean the median level? The mean? The root mean square? Or maybe the minimum level? Such ambiguity could kill.

But the safety cases for nuclear plants are 鈥渧ery, very, very technical documents鈥�, says an inspector from the Nuclear Installations Inspectorate, who did not wish to be named. He defies any expert in literary criticism to analyse them: 鈥淵ou鈥檇 need the right background in nuclear engineering to do it.鈥�