Understanding and modelling human error for interactive system design

Speaker: Ann Blandford

11th January 2006 , 2pm , Devonshire G21/G22 Conference Room

Abstract

Most work on human error is based around incident reports or other forms of retrospective reporting. This has led to important insights into causes of error – e.g. on the roles of context, feedback and policies in provoking human error. However, this approach cannot give insights into many of the underlying cognitive causes of error – what Hollnagel calls the ‘genotypes’ of error. One of the great challenges of conducting controlled studies of error is that they are difficult to provoke systematically in a controlled laboratory setting. In this talk I will present our ongoing work on developing methods for studying error in controlled settings that enable the manipulation of variables such as interruption point, task structure and visual cueing to determine their influence on certain types of error. In particular, we have been studying the roles of memory and attention in omissions such as post-completion errors. This work is informing parallel studies on formal modelling of users for systems verification, and less formal work on error evaluation methods.