Dynamic detection of novice vs. skilled use without a task model

  • Authors:
  • Amy Hurst;Scott E. Hudson;Jennifer Mankoff

  • Affiliations:
  • Carnegie Mellon, Pittsburgh, PA;Carnegie Mellon, Pittsburgh, PA;Carnegie Mellon, Pittsburgh, PA

  • Venue:
  • Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
  • Year:
  • 2007

Quantified Score

Hi-index 0.01

Visualization

Abstract

If applications were able to detect a user's expertise, then software could automatically adapt to better match exper-tise. Detecting expertise is difficult because a user's skill changes as the user interacts with an application and differs across applications. This means that expertise must be sensed dynamically, continuously, and unobtrusively so as not to burden the user. We present an approach to this prob-lem that can operate without a task model based on low-level mouse and menu data which can typically be sensed across applications at the operating systems level. We have implemented and trained a classifier that can detect "nov-ice" or "skilled" use of an image editing program, the GNU Image Manipulation Program (GIMP), at 91% accuracy, and tested it against real use. In particular, we developed and tested a prototype application that gives the user dy-namic application information that differs depending on her performance.