An Open Source Approach to Information Scent

Abstract

Several theories of how novices explore a new interface have arisen in HCI and have led to methods and tools for UI practitioners to predict users’ behavior on proposed interfaces and improve their design ideas before implementation. These tools depend on obtaining quantitative estimates of information scent, which usually requires a large corpus of documents representing the background knowledge of target users and an algorithm that uses this corpus to calculate the information scent between a user’s goal and the labels in the GUI. Prior work has often used proprietary algorithms and propriety, now outdated corpora. This note presents an open source approach to information scent that is as good a foundation for research and practice as a previously published proprietary system, exemplifies its use in CogTool-Explorer, and opens the door for researchers to explore many important questions about novice exploration of interfaces.