Reference information
Author: Dustin A. Smith, Henry Lieberman
When/Where: IUI’10, February 7–10, 2010, Hong Kong, China.
Summary
This paper explores the incorporation of goal networks into user interfaces. When an application is aware of a user’s goal, knowledge about the relationships between goals is useful for automated planning, to coordinate workflows that span multiple interfaces and to build introspective applications. The designers gather goals and goal steps, or subtasks, from a community using natural language processing and plan recognition to create a goal network.
Goal libraries are used to map interface-level goals to human-level goals (goals that people communicate with each other). The designers took data from 43Things.com, a website where users share and discuss goals and plans in order to acquire data for their goal library. First, they took people's goal statements to implement a directed, weighted goal graph. This graph was clustered to trim the graph and eliminate outliers, then verb-phrases were extracted and associated with rules. The human goals were analyzed to find which were most popular, controversial, and which goals were most central to the entire goal network. This goal knowledge was then demonstrated in a simple interface called ToDoGo.
Discussion
The focus of this paper is really more on the concept of goal extraction from natural language. The relation to interface design was rather vague, and I struggle to see how it fits. The example application itself really took a backseat in the paper. There could have been a small preliminary user study to test its efficacy. Still, as an idea, I believe there is a lot of promise, and I'm curious to see how it can influence interface design. I liked that there was discussion of certain artificial intelligence (AI) concepts, particularly search, natural language processing and goal-based agents.
No comments:
Post a Comment