Paper Reading #24 - The Why UI: Using Goal Networks to Improve User Interfaces

Comments:


Reference:
The Why UI: Using Goal Networks to Improve User Interfaces
Dustin A. Smith, Henry Lieberman
IUI '10

Summary:
This paper examines a new method of modeling user goals through voice commands.  The method the authors describe processes a natural language plan corpus, constructed by many people.  Goals can be inferred by a system or explicitly stated by a user.  However, each goal can have a multitude of sub-goals, and it is useful to predict specifically what the user is interested in.  As it is, people usually fall short in stating explicit goals. 

The authors downloaded a corpus from 43things.com which contains a multitude of goal statements ("buy a house," "travel to New York," etc.)  These goal statements did not have explicit relations, so the authors attempted to infer relations by matching the context of plan statements with goal statements.  After correcting spelling mistakes and in general tidying up the text, they built a directed weighted goal graph.  They found that while most of the goals in the corpus were unique, there were a few that were far more prevelant than others.  They then transfer the results to a mobile app which can help the user achieve those goals.  For example, a user wanting to buy a house would have a map come up suggesting subgoals, such as locations and details for realtors in the area.

Discussion:
This is another interesting example of how technology may be used to assist us in everyday life.  This application doesn't do anything that people with the time and motivation couldn't figure out by themselves, but it certainly makes the potentially desired information easier to access.  I can especially see it being useful for someone trying something for the first time.  It will be interesting to see where this technology will go from here.