Resolving Complexity in Accident Texts Through Graphical Notations and Hypertext

Accident reports are the result of investigations into accidents that have occurred in safety critical systems. It has been suggested that accident reports can be both complex and difficult texts to understand resulting in problems of comprehension. Graphical notations have been suggested as a way to resolve these difficulties due to their concise and visual nature. However, there is no evidence in the literature to support these claims. This thesis investigates the value of using graphical representations and hypertext to help resolve complexity in accident texts, and thus, aid their comprehension. The first part of the thesis explores whether graphical representations have a role to play in aiding comprehension of accident texts. We show that accident texts are complex in their representation of event sequences. Three graphical notations are introduced Petri Nets, Accident Fault Tress and Why-Because Graphs. First, we analyse their properties to evaluate how accurately they represent event sequences. Then we use comprehension experiments to measure if they do indeed aid understanding. Finally, we observe the process of abstraction to determine the effects on comprehension. Although the findings suggest that graphical representations do help, there is strong evidence for the need for multiple representations. Subsequently, a hypertext tool is envisaged to help accident analysis using both text and graphical representations. Based on this conclusion, the second half of the thesis evaluates the feasibility of this concept through industrial research with BAE SYSTEMS using HCI techniques. A paper prototype interface is designed and evaluated further by co-operative evaluations with accident report users from BAE SYSTEMS and the University of York.