Next: Manual Browsing and Up: The Graphical User Previous: Reading News Retrieved

Providing Feedback for Articles Retrieved

The user can provide feedback for the articles retrieved by the agent. Under each Agent Icon are small icons bearing the signs ``+'' and ``-'' to enable the agent to receive positive and negative feedback respectively. Clicking on one of these provides the appropriate feedback to the agent for the current article selection. ``Current Selection'' refers to the article that was last selected for reading in any of the News Windows. There is only one current article selection across all News Windows. This is so that any agent can be provided with feedback for any article (especially required in the case of manual browsing and programming by demonstration, see Section ).

When feedback is provided as described above, the implicit assumption is that the feedback is for all the features of the article i.e. the keywords that were used to decide that the article should be retrieved, as well as the other fields in the article. All are used to modify the profile in consonance with the feedback. If the user wants to selectively provide feedback for a certain feature in the article (as opposed to the article as a whole), the user can highlight that portion of the text which contains the feature and then provide positive or negative feedback. For example, the user can highlight a paragraph in the text of the article and provide positive feedback. In this case, only the terms in the paragraph will be added to the profile instead of terms in the entire article.

The user can also seek explanations for selecting the current document. By clicking on the icon bearing the sign ``?'', the user can ask an agent why it recommended the current article selection. If the agent retrieved the document, it presents a brief explanation for recommending the document. In the current implementation, the only information provided as part of the explanation is the document score. Future versions are expected to have more detailed explanations.



Next: Manual Browsing and Up: The Graphical User Previous: Reading News Retrieved


MIT Media Lab - Autonomous Agents Group - agentmaster@media.mit.edu