Enhanced task modelling for systematic identification and explicit representation of human errors

Abstract

Task models produced from task analysis, are a very important element of UCD approaches as they provide support for describing users goals and users activities, allowing human factors specialists to ensure and assess the effectiveness of interactive applications. As user errors are not part of a user goal they are usually omitted from tasks descriptions. However, in the field of Human Reliability Assessment, task descriptions (including task models) are central artefacts for the analysis of human errors. Several methods (such as HET, CREAM and HERT) require task models in order to systematically analyze all the potential errors and deviations that may occur. However, during this systematic analysis, potential human errors are gathered and recorded separately and not connected to the task models. Such non integration brings issues such as completeness (i.e. ensuring that all the potential human errors have been identified) or combined errors identification (i.e. identifying deviations resulting from a combination of errors). We argue that representing human errors explicitly and systematically within task models contributes to the design and evaluation of error-tolerant interactive system. However, as demonstrated in the paper, existing task modeling notations, even those used in the methods mentioned above, do not have a sufficient expressive power to allow systematic and precise description of potential human errors. Based on the analysis of existing human error classifications, we propose several extensions to existing task modelling techniques to represent explicitly all the types of human error and to support their systematic task-based identification. These extensions are integrated within the tool-supported notation called HAMSTERS and are illustrated on a case study from the avionics domain.