Interface Style and Error Rates - Some Experimental Results

Abstract

At the IPP a number of new forms of visualizations of process values have been developed. Several of them have been evaluated by user experiments. In the context of a research project (supported by the Volkswagen Foundation in Germany (Ref. Nr.: U69 886)) a comparative study with respect to the influence of the interface style on the error rate during classification of process states was undertaken. The paper describes the results and discusses them in a taxonomy context. NEW FORMS OF VISUALIZATIONS:"During the past years at the IPP a number of display types have been developed with the aim to reduce the time for detection of undesired process states as well as the error rate of operators. Two of them have were extensively evaluated by means of user experiments. The 'Mass Data Display' (MDD)This type of display' (cf. Figure 1) was developed in cooperation between IPP and ABB and first presented at the Interkama trade fair in 1992 (Zinser, 1993) The goal of this development was to identify a technique that enables the operator to maintain an overview over the entire process with moderate effort and thus to develop a "feeling" for the overall behaviour of the process and upcoming abnormal situations.Technical principle. The approach is mainly based upon the work of Grinstein (Grinstein and Bergeron, 1989). The design principle is that each process value is represented by a small icon on a display screen. This icon changes its shape (size, colour etc) according to the changes of the process value. The operator does not observe the behaviour of each individual icon, but the overall pattern that develops because a number of icons -that belong to technologically interrelated process entities change their appearance simultaneously. This technique allows to show an overview over several hundreds (even thousands) of values on one display screen of average size, Experiments. In order to assess the validity and usability of the design extensive experiments were undertaken. Three abnormal situations were presented to test persons and the time for the correct identification of these situations measured. As a cross-check a fourth situation was presented to the test persons: a load transition of +I-10 90. This was chosen because it produced changes of the display pattern that were similar to that of abnormal status. The test persons were students of the TUC. There were three groups of four students with different levels of knowledge about technical processes: beginners, persons with basic knowledge and well trained persons. 1 I.000 individual measurements were collected. The results were evaluated wine a commercial software nackaee for stati-L u stical analyses.Results. It turned out that all test persons identified the abnormal situations much earlier when using the MDD in contrast to the P&I diagrams. In one case they were even 20 times faster. This result even exceeded the expectations of the authors. Figure 2 shows a simplified plot of the results in two situations. The experiments and their results have been described in d...