Agents that reduce work and information overload

Publication Type: 
Journal Article
Authors: 
Pattie Maes
Journal: 
Commun. ACM
Publisher: 
ACM Press
Volume: 
37
Number: 
7
Pages: 
30-40
Year: 
1 994
Place Published: 
New York, NY, USA
URL: 
http://web.media.mit.edu/~pattie/CACM-94/CACM-94.p1.html
Abstract: 
The "information highway" will present us with an explosion of new computer-based tasks and services, but the complexity of this new environment will demand a new style of human-computer interaction, where the computer becomes an intelligent, active and personalized collaborator. Interface agents are computer programs that employ Artificial Intelligence techniques to provide active assistance to a user with computer-based tasks. Agents radically change the current user experience, through the metaphor that an agent can act as a personal assistant. The agent acquires its competence by learning from the user as well as from agents assisting other users. Several prototype agents have been built using this technique, including agents that provide personalized assistance with meeting scheduling, electronic mail handling, electronic news filtering and selection of entertainment.
Notes: 

Employing agents fo delegate certain computer-based tasks such as dealing with junk mail, scheduling and rescheduling meetings, searching for relevant information among heaps of irrelevant information, and browsing through lists of books, music, and television programs in search of something interesting.
News filtering agent: helps the user select articles from a continuous stream of news. The agent is initialized by giving it some positive and negative examples of articles to be retrieved. The agent performs a full text analysis (using the vector-space model for documents) to retrieve the words in the text that may be relevant. It also remembers the structured information about the article, such as the author, source, assigned indices, and so forth. Once an agent has been bootstrapped, it will start recommending articles to the user. The user can give it positive or negative feedback for articles or portions of the article recommended. The user can also select the author or source and give positive and negative feedback. This will increase or decrease the probability that the agent will recommend similar articles in the future. It performs content filtering, in the sense that the agent tries to correlate the positive and negative feedback with the contents of the article.
Entertainment agent: recommends books, music or other forms of entertainment. They use social filtering, i.e. they do not attempt to correlate the user's interests with the contents of the items recommended. Instead, they rely solely on the correlations between different users.
Users will need help dealing the information and work overload. Users can delegate a range of tasks to personalized agents that can act on the user's behalf. Agents gradually learns how to better assist the user by:

- Observing and imitating the user;
- Receiving positive and negative feedback from the user;
- Receiving explicit instructions from the user.