The Effect of Interface Elements on Transcription Tasks to Reduce Number-Entry Errors

Many tasks in daily life require transcribing information accurately from one medium to the other. However, humans make errors frequently. While most of the errors in daily life are little more than an inconvenience in safety- critical domains, such as healthcare, a small error like typing the wrong number when programming a medical device can have grave consequences. Despite potentially fatal consequences of errors, little is known about the errors people make using medical devices, such as infusion pumps, and how the devices themselves influence the errors made. This thesis reports ten studies looking at different interface design features in the context of medical devices and their potential influence on reducing errors.
The first three studies empirically evaluated particular design features pro- posed by previous work. These studies did not produce the predicted error reduction although there was a recurrent low rate of errors in all the studies. Therefore such claims made in earlier work could not be supported. The studies however showed that it is hard to see the impact of particular features due to the nature of the research area and due to the robustness of the evaluated interface. Additionally, interesting insights into how people use such interfaces to enter numbers need to be taken into account.
Inspired by results from cognitive psychology, which suggest that representing information in a poorer quality format increases the likelihood of memorising the information more accurately, a further set of seven experiments are presented in this work evaluating the effect of such an approach on number transcription tasks. Results showed that people made significantly fewer errors when transcribing less visible numbers as well as text. Furthermore, the studies also confirmed that the source displaying the number and not the entry display is responsible for this counter-intuitive approach to reduce errors. This aligns with previous work in human-computer inter- action and psychology. Moreover, the robustness of the discovered effect was investigated once levels of audio distractions were present. Even in a distracting environment the effect led to a significant decrease in errors.
The potential impact of this work could be a valuable contribution for do- mains where accuracy is of great importance.