[Topic 1] Human Agent Interaction based on Instant Messenger and its Application on Electronic Commerce

In this Topic, we develop a collaboration mechanism among human and software agents based on instant messengers, e.g. AOL Instant Messenger (AIM), MSN Messenger, Yahoo! Messenger, iChat, etc. As an application, we construct a bidding support system for Internet Auctions.

Instant Messengers (IM) have been used very widely. The number of their users exceeds 3 million in Japan. Especially, in AIM, near 2 billion messages are exchanged every day. The main functions of IM are exchanging messages and monitoring current situations of members. The advantage of IM is its easy and informal nature for communication. In our group, we employ IMs for user interface between users and agents. IM can facilitate informal collaboration between users and software agents.

Technical issues we needed to concern are the following:

    Issue 1: To communicate with the existing standard protocols (AIM, MSN Messenger, Yahoo! Messenger, etc.).
    Issue 2: To design commands for software agents.
For these issues, we employ the Jabber technology. The Jabber technology enables us to make programs that can communicate XML-based messages based on the existing standard protocols.

Further, we focus on an important issue that related to the trade-off between speedy decision making by software agents and high-quality decision making by human. Software agents can make a decision quickly under dynamic environments. But, their decision may become a low quality decision. On the other hand, human can make a high quality decision compared with software agents. But, it can take longer time than software agents. Especially, under dynamic environments, in one case, human needs to make a high-quality decision. In the other case, agents need to make decision quickly. In this work, we develop a mechanism that can adjust agent's autonomy by using dynamic probabilistic network. As a real application, we are implementing a bidding and goods-monitoring system for electronic commerce.


Shintani lab. 2003