ABSTRACT
We are all familiar with the annoyance of errors that we make with everyday devices , such as switching on an empty kettle , or making mistakes in the programming sequence with a video cassette recorder. People have a tendency to blame themselves for 'human error' . However, the use and abuse of the term has led to authors questioning the very notion of human error (Wagenaar and Groeneweg, 1988 ) . 'Human error' is often invoked in the absence of technological explanations. One can argue that human error is not a simple matter of one individual making one mistake, so much as the product of a design which has permitted the existence and continuation of specific activities which could lead to errors (Reason, 1 990) . Predicting human error may strike the reader, at first sight, as implausible. However, if we know an activity that is to be performed, and the characteristics of the product being used, then it should be possible to indicate the types of error which may arise. This is the gist of all methods aimed at predicting human error: first define what actions need to be performed , and then indicate how these actions might fail. Techniques have been developed for the detailed and systematic assessment of a person 's activities . A structured approach enables an analyst to identify potential points in tasks where errors could have significant negative consequences . From this assessment, preventive strategies can be sought to minimise the consequences or reduce the likelihood of error. An abundance of methods for identifying human error exist, some of which may be appropriate for the analysis of consumer products. In general , most of the existing
techniques have two key problems . The first of these problems relates t o the lack of representation of the external environment or objects . Typically , human error analysis techniques do not represent the activity of the device and material that the human interacts with, in more than a passing manner. Hollnagel ( 1 993 ) emphasises that human reliability analysis (HRA) often fails to take adequate account of the context in which performance occurs. The second problem of existing techniques is that there tends to be a good deal of dependence made upon the judgement of the analyst. Different analysts , with different experience , may make different predictions regarding the same problem (caned inter-analyst reliability) . Similarly , the same analyst may make different judgements on different occasions (intra-analyst reliabil ity ) . This subjectivity of analysis weakens the confidence that can be placed in any predictions made.
