Modeling strategic use of human computer interfaces with novel hidden Markov models

Area of Science:

• Cognitive Science
• Human-Computer Interaction
• Data Science

Background:

• Immersive software tools create virtual environments for data interaction and manipulation.
• User interactions within these environments generate detailed activity logs, providing context for cognitive and behavioral processes.
• Analyzing these logs offers a high-resolution view of dynamic user behaviors.

Purpose of the Study:

• To introduce novel methods for analyzing and interpreting data from immersive software environments.
• To validate a new modeling approach using a preliminary study involving user learning in a game.
• To understand how users learn to integrate software functionality for strategic task pursuit.

Main Methods:

• Utilized a novel implementation of the Beta Process Hidden Markov Model (BP-HMM) for analyzing software activity logs.
• Collected interaction data from 20 participants learning a new computer game.
• Jointly modeled activity log sequences using BP-HMM to identify collective behavioral patterns.

Main Results:

• Identified a global library of activity patterns representative of collective user behavior.
• Found systematic relationships between pre- and post-task questionnaires (problem-solving styles, engagement, workload) and BP-HMM metrics.
• Demonstrated that the BP-HMM approach can decompose unstructured behavioral data effectively.

Conclusions:

• The novel BP-HMM approach provides a sensible method for understanding user learning and skill acquisition in software environments.
• This method allows for the analysis of dynamic cognitive and behavioral processes through software activity logs.
• The findings suggest a way to link self-reported problem-solving styles to observable user behavior within software.