Edinburgh Research Explorer

Using Web Site Synthesis in an Experiment on the Causal Perception of Aviation Accidents

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Original languageEnglish
Title of host publicationWorkshop on the Investigation and Reporting of Incidents and Accidents (IRIA 2002)
Place of PublicationGlasgow, UK
PublisherDepartment of Computing Science, University of Glasgow
Pages221-230
Number of pages10
Publication statusPublished - 2002

Abstract

An obvious way of presenting incident and accident information to the public is via Web sites. There is, however, little understanding of how best to present this sort of material. We do not even know how sensitive people might be to differences in the way we construct our Web sites. A key issue is the degree to which those viewing an air accident Web site believe that certain events cause accidents. There may be certain styles of presentation or navigation within and between web-based accident reports that can either hinder or support the readers’ ability to interpret evidence about previous failures. This, in turn, can undermine the argument that investigators present to support the causes that they distinguish in their report. The psychological literature offers various models of causal perception. If a model could be found which (even roughly) predicted strength of causal perception for incident/accident reporting Web sites then this
would help us design sites which are more likely to give the perceptions of causality which we intend. The problem is that the available predictive models do not give similar predictions. We describe an experiment investigating the predictive power of two models of causal perception applied to accident reporting Web sites. Our experiment is novel in its use of automated synthesis to construct experimental Web sites which are guaranteed to have the same content but which vary in Web site structure according to a number of key parameters. This allows us quickly to construct experiments involving large but closely comparable Web sites.

ID: 3505593