Date: Tue, 10 Dec 1996 21:39:56 GMT Server: NCSA/1.4.2 Content-type: text/html Last-modified: Fri, 17 Mar 1995 21:58:40 GMT Content-length: 5636
This document continues recent discussions on the User Interface for an Intelligent Agent that assists in using the World Wide Web.
Finding (and using) resources on a large, loosely-organized network can be tedious and time-consuming. Intelligent agent software can assist a user, but current user interfaces are either hard to use or do not fully exploit the agent's capabilities.
I want to construct a testbed for evaluating different user interfaces for assisted resource discovery. The audience consists of casual or "naive" users of information networks (i.e. I don't expect them to be able to program or to know predicate logic).
What are the User Interface research issues that I want to address?
Is providing a software assistant a natural and easy metaphor for generating fruitful searches of large, loosely-organized networks?
What user interface elements of an assisted search contribute to a rapid convergence on a result? How can the user work with the assistant to initially specify, then modify the goals of the search?
Is it natural to classify types of resource discovery tasks? If so, are there specialized user interface elements that are appropriate for these classifications?
Once many resources are discovered, how can they be retained and managed by the user for future access? What user interface metaphors facilitate effective resource management? Such metaphors should integrate with the assistant (e.g. an assistant should keep the resource location current).
This work will use a software robot that can understand and navigate the World Wide Web. This is not the only domain. It is a very well-known domain in which easy, frutiful resource discovery is crucial. In the World Wide Web domain, a most common example of resource discovery is Search.
In general this approach applies to all types of resources and to information network services other than discovery. Ultimately I would like to see the most effective user interface techniques tightly integrated into a Global Desktop that has built-in software assistants to accomplish many chores we would normally want to delegate to human assistants.
As a start I will work on adding an assistant-based interface for resource discovery on top of the Mosaic World Wide Web browser. I have considered two scenarios:
Scenario 1 would provide a companion application to Mosaic that allows a user to search for Web resources. Initially this would implement a keyword search but it would be built with the idea of adding specialized searches later. The search is carried out by a software agent (so I presume the UI generates a goal that the agent tries to satisfy) which returns a collection of Web resources. The user can then tour the results by directing the companion to control Mosaic remotely. While the search is in progress, the user can continue to work with Mosaic.
This approach would provided the same search mechanisms as Scenario 1. The UI would be integrated into the Web as another Web page (like Webcrawler or Yahoo). It would also build a keyword-based query for a software agent, but using Mosaic forms. The results would be returned in a generated HTML document and the standard hypertext navigation would be used to visit the results.