Automatic Chunk Detection in Human-Computer Interaction
Santos, Paulo J.
MetadataShow full item record
This paper describes an algorithm to detect users' mental chunks by analysis of pause lengths in goal-directed human-computer interaction. Identifying and characterizing users' chunks can help in gauging the users' level of expertise. The algorithm described in this paper works with information collected by an automatic logging mechanism. Therefore, it is applicable to situations in which no human intervention is required to perform the analysis, such as adaptive interfaces. An empirical study was conducted to validate the algorithm, showing that mental chunks and their characteristics can indeed be inferred from an analysis of human-computer interaction logs. Users performing a variety of goal-directed tasks were monitored. Using an automated logging tool, every command invoked, every operation performed with the input devices, as well as all system responses were recorded. Analysis of the interaction logs was performed by a program that implements a chunk detection algorithm that looks at command sequences and timings. The results support the hypothesis that a significant number of user mental chunks can be detected by our algorithm.