Glasgow Accident Analysis Group

News: Rebecca Randell was awarded the Diane Forsythe prize 2004, for a paper on the customisation of medical devices. This was written while she was a member of GAAG and the prize is awarded annually by the American Medical Informatics Association.

The Glasgow Accident Analysis Group is a small team of researchers based in Glasgow University. Our aim is to improve our understanding of system `failures' and human `error'. We are interested in accident analysis and in incident reporting techniques. There are several strands to our work:

new techniques for incident reporting. Our work focusses on two ends of the spectrum. For local incident reporting systems, we are looking at ways in which workplace studies can be used to uncover the deeper causes that can lie behind individual incidents that are reported to these systems. For national schemes, we are looking at the automated support (case based reasoning and information retrieval tools) that is necessary to identify common causes amongst many thousands of incident reports. We are also particularly interested in the reporting and analysis of adverse events involving programmable systems. This work has been support by the Health and Safety Executive, the UK equivalent of OSHA and by joint projects to develop guidelines on incident reporting within European air space with EUROCONTROL. Novel Computational Techniques for Incident Reporting , Classification and Analysis of Incidents in Complex, Medical Environments, Novel Techniques for Analysing Computational System Failure
the pragmatic application of analytical methods. We are using formal and semi-formal notations not simply to design safety-critical applications, but also to identify the regulatory, managerial, human and systems failures that lead to major accidents. Much of this work has been conducted with C.Michael Holloway at NASA Langley. Some notes on Proving properties of accidents and Visualizing the Relationship between Human Error and Organizational Failure .
the pragmatic integration of cognitive models and human error models. Existing taxonomies of human error are often pitched at too high a level to support subsequent re-design. We are using cognitive models in conjuction with these taxonomies to represent and reason about the deeper causes of operator 'failure'. Here are some notes on Why human error analysis fails to support systems development
Observational methods for understanding technologies in use. While accident reports can help us understand problems after an event, we use observational methods to locate problems with technologies before an accident occurs. Much of this work has been conducted within the context of the UK National Health Service. Other projects have looked at UK railways and several military organisations around the globe. For example, while adaptation of devices and normalisation of problems have been cited as contributory factors to accidents, our studies look at how these behaviours come about and why they persist. Coping Strategies and User Adaptation of Medical Devices
the use of leading edge visualisation techniques. We are trying to go beyond the use of fault trees and wire-frame simulations of major accidents. We are using imagemaps, photorealistic simulations and model based virtual reality to improve the visualisation of major failures using standard PCs and web browsers. Here are some notes on Improving the Presentation of Accident Reports over the World Wide Web. Some related work on 3D audio for alarms.

For further information about this group and its activities, please contact:

Prof. Chris Johnson,
Dept of Computer Science,
University of Glasgow,
Glasgow,
G12 8QQ,
Scotland.

Phone: +44 141 330 6053
Fax:   +44 141 330 4913
E.mail:  johnson@dcs.gla.ac.uk