Effects of Interpretation Error on User Learning in Novel Input Mechanisms

Abstract

Novel input mechanisms generate signals that are interpreted as commands in computer systems. Sometimes noise from various sources can cause the system to produce errors when attempting to interpret the signal, causing a misrepresentation of the user's intention. While research has been done in understanding how these interpretation errors affect the performance of users of novel signal-based input mechanisms, such as a brain-computer interface (BCI), there is a lack of knowledge in how user learning is affected. Previous literature in command-based selection tasks has suggested that errors will have a negative impact on expertise development; however, the presence of errors could conversely improve a user's learning by demanding more attention from the user. This thesis begins by studying people's ability to use a novel input mechanism with a noisy input signal: a motor imagery BCI. By converting a user's brain signals into computer commands, a user could complete selection tasks using imagined movement. However, the high degree of interpretation errors caused by noise in the input signals made it difficult to differentiate the user's intent from the noise. As such, the results of the BCI study served as motivation to test the effects of interpretation errors on user learning. Two studies were conducted to determine how user performance and learning were affected by different rates of interpretation error in a novel input mechanism. The results from these two studies showed that interpretation errors led to slower task completion times, lower accuracy in memory recall, greater rates of user errors, and increased frustration. This new knowledge about the effects of interpretation errors can contribute to better design of input mechanisms and training programs for novel input systems.